1.0.json

{
  "openapi": "3.2.0",
  "info": {
    "title": "DiffScope Synthesis Platform API",
    "version": "1.0",
    "description": "DSSP HTTP API surface represented by this service. A server implementation can support a subset of the broader DSSP API.\n\n## Streaming responses\n\n### Response negotiation\n\nOperations whose response content declares application/x-ndjson support streaming on the same URL with the same method and request body as the corresponding non-streaming operation. The Accept request header specifies acceptable response media types. Accept: application/x-ndjson requests a stream. Clients that can process NDJSON responses should prefer this media type when specifying Accept. Accept: application/json requests a non-streaming response. If both media types are acceptable, selection follows HTTP content negotiation, including quality values. If Accept is absent or contains only */*, the response uses application/json. The response Content-Type identifies the selected representation. Request bodies retain Content-Type: application/json.\n\n### NDJSON response format\n\nAn application/x-ndjson response is a UTF-8 sequence of JSON objects. Each object occupies a single line terminated by LF. Line breaks within JSON strings are escaped. Each line contains one event with the envelope {\"event\": \"<event name>\", \"data\": <event data>}. The data member is omitted for queue and is optional for progress. Each operation's itemSchema describes one event. Event ordering is governed by the following table and is not validated by itemSchema.\n\n| Event | Data | Meaning | Allowed next events |\n| --- | --- | --- | --- |\n| queue | Absent. | The task is queued on the server. Disconnecting in this state does not discard any unfinished processing progress. | queue, progress, result, error |\n| progress | Optional object {\"value\": <number from 0 to 1 inclusive>}. Absent data is equivalent to {\"value\": 0}. | The task is processing. The normalized value is solely for client progress display and has no business meaning. | progress, result, error |\n| result | Required. The complete response body defined for application/json by the same operation. | The task completed successfully. Results are returned in full in a single event. | None. This event is terminal. |\n| error | Required. The error data format is unspecified, as for non-streaming errors. | The task failed. | None. This event is terminal. |\n\nThe stream may begin with any event. Queue and progress events are optional. A completed stream contains exactly one terminal event, result or error, and no events follow it. A connection that closes before a terminal event does not establish task completion. A progress value of 1 does not replace a result event.\n\nA negotiated application/x-ndjson response always uses HTTP status 200, including if its first event is error. Request decoding, validation, architecture selection, task execution, and result encoding failures use an error event. The error data contains the corresponding problem response body. The status member in that body describes the failure and does not change the HTTP response status. If neither application/json nor application/x-ndjson is acceptable, the server returns HTTP 406.\n\nDuring queue and progress, the server repeats the current event at the configured heartbeat interval, including if the progress value is unchanged. The default interval is 10 seconds. Preparation does not produce progress events. Only the initial resource admission produces queue events. Subsequent resource waits retain the last progress state. Progress measures the task body from 0 to 1 and excludes preparation and final result construction or encoding. A value of 1 indicates completion of the task body. A later error event remains possible. Tasks without model work may return a terminal event directly.\n\nExample of a successful /synth/duration response:\n\n```ndjson\n{\"event\":\"queue\"}\n{\"event\":\"progress\"}\n{\"event\":\"progress\",\"data\":{\"value\":1}}\n{\"event\":\"result\",\"data\":{\"output\":{\"notes\":[]}}}\n```\n\n## Synthesis workflow\n\nSynthesis proceeds from score data through pronunciation, phonemes, phoneme timing, parameter curves, and encoded audio. Clients select stages from the chosen group's capabilities and supply existing or manually prepared data for stages that do not require automatic generation. Each request supplies its context and the complete input required by that stage. Clients retain the score and intermediate results and assemble subsequent requests from them.\n\n### Discovery and context\n\n1. Select an architecture through `GET /arch` and a singer through `GET /singer` or the architecture's singer endpoints. Use the singer's `arch` and `group` to obtain the corresponding group metadata through `GET /arch/{arch_id}/group/{group_id}`. Architecture metadata defines parameter properties. Group metadata defines available singing languages, conversion modes, duration prediction, parameter dependencies, audio dependencies, and mixing support.\n\n2. Construct `context.arch` from the selected architecture. Initialize `context.archExtra` from the architecture's `defaultArchExtra` and each singer source's `extra` from that singer's `defaultExtra`, applying any requested architecture-specific settings. Pronunciation and phoneme requests use `SingleSingerContext`. Duration, parameter, and audio requests use `MultiSingerContext`, including for a single source. Multiple sources must share an architecture and group, and the group must have `mixable=true`. Keep the source order consistent with the per-frame `mix` data.\n\n3. Select each note's singing language from the group's `languages` map. The singer's nonempty `defaultLanguage` provides a preferred language. `Accept-Language` selects metadata display text and does not select the singing language. `POST /synth/env-tag` returns an environment tag for the supplied synthesis context.\n\n### Pronunciation and phonemes\n\n1. For a language with `pronunciationMode=full`, call `POST /synth/pronunciation` with each note's lyric and language. Retain the returned pronunciation or select an alternative from its candidates. For `pronunciationMode=skip`, supply pronunciation directly. Existing pronunciations may also be used without conversion.\n\n2. For `phonemeMode=full` or `token_only`, call `POST /synth/phoneme` with the selected complete pronunciation and language for each note. Optional `syllableSliceStart` and `syllableSliceEnd` allocate syllable groups to a note. `full` supports tokens and syllable onset labels. `token_only` supports token conversion without onset labeling, so clients must supply any onset labels required by later stages. For `phonemeMode=skip`, supply phonemes directly.\n\n3. Assemble score notes with position, pitch in cents, pronunciation, language, allocated phonemes, and `kind`. Copy returned phoneme tokens and onset flags and add the appropriate phoneme language. Express a slur with `kind=slur`, an empty phoneme array, and zero gap. A slur cannot be the first note. Each note retains its own complete pronunciation and language. Pronunciation strings do not encode slurs or syllable allocation.\n\n### Phoneme timing\n\nPrepare `pieceDuration`, score notes, `mix`, and `mixSampleRate` for the piece. If the group has `durationMode=full`, call `POST /synth/duration`. Merge each returned `output.notes[i].phonemes[j].start` into the corresponding input phoneme, retaining the score fields, tokens, onset flags, and languages. The resulting timed notes are the input to parameter and audio generation. If `durationMode=skip`, supply phoneme starts directly. Existing or manually edited starts may also be used without prediction.\n\n### Parameter preparation\n\nStart with the group's `audioDependencies` to identify the curves required for audio generation. For every curve to be predicted, follow its `dependsOn` entries in `parameterPipeline` recursively. Prepare dependencies before dependent predictions. Parameters with `type=direct` are supplied by the client. Parameters with `type=indirect` may be supplied as existing curves or generated through `POST /synth/parameter`.\n\nEach parameter request supplies the timed notes, piece duration, mix data, and parameter curves with their sampling rates. Add `retake` only to parameters that require prediction. Supply dependency curves without `retake` to use them as conditions. Their presence does not request regeneration. A nonempty retake range uses frames on that parameter's input sampling grid. `retakeMode=full` requires the complete piece. `retakeMode=range` also permits subranges. A zero-length range performs no recomputation, and direct parameters cannot have a nonempty retake range.\n\nRetain the returned `output.parameters` curves and their returned sampling rates for subsequent requests. For example, if pitch depends on direct expressiveness and energy depends on pitch, supply expressiveness, predict pitch with `retake`, then supply the returned pitch without `retake` when predicting energy. If pitch or energy is already available, supply that curve directly and omit its prediction. After an edit, refresh affected timing and dependent curves before regenerating audio.\n\n### Audio generation\n\nCall `POST /synth/audio` with the prepared timed notes, piece duration, mix data, and all curves required by `audioDependencies`. Audio parameter entries contain values and sampling rates without `retake`. Audio generation does not implicitly predict missing parameters. Use `acceptableFormats` and `acceptableSchemes` to constrain the encoded audio representation and URL scheme. Retrieve or decode the audio identified by `output.audioUrl` according to its scheme.\n\nFor operations that declare NDJSON responses, the response negotiation and event rules apply to each stage independently. Consume a successful terminal `result` event before using its data as the next stage's input.\n\n### Long content and segmentation\n\nFor long synthesis content, clients should divide the score into segments at blank intervals between phrases and synthesize each segment separately. Choose boundaries within these intervals without splitting a sustained note or slur. Reserve a short leading margin before the first note for consonant onsets and a short trailing margin after the last note for vocal tails. The margins should accommodate the phoneme timing and vocal decay rather than end exactly at note boundaries.\n\nInclude both margins in each segment's `pieceDuration` and prepare its note timing, mix data, parameter curves, and retake ranges for the segment's local timeline. Retain the segment's offset in the original score. Place the generated audio at that offset, including the leading margin, so the notes remain aligned with the score. Preserve the original blank intervals and avoid trimming audible consonant onsets or vocal tails during assembly."
  },
  "servers": [
    {
      "url": "http://localhost:13711/v1",
      "description": "Default local server URL. The actual host and port are configured by the service."
    }
  ],
  "tags": [
    {
      "name": "Application",
      "description": "Application metadata endpoints."
    },
    {
      "name": "Synthesis",
      "description": "Synthesis service endpoints."
    },
    {
      "name": "Extraction",
      "description": "Audio analysis and extraction endpoints."
    }
  ],
  "paths": {
    "/info": {
      "get": {
        "tags": [
          "Application"
        ],
        "operationId": "getApplicationInfo",
        "summary": "Get application information",
        "description": "Returns service application metadata.",
        "responses": {
          "200": {
            "description": "Application information.",
            "content": {
              "application/json": {
                "schema": {
                  "$ref": "#/components/schemas/ApplicationInfoResponse"
                }
              }
            }
          }
        }
      }
    },
    "/arch": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchitectureList",
        "summary": "List architectures",
        "description": "Returns metadata for every registered architecture. Display language is specified via the Accept-Language request header.",
        "responses": {
          "200": {
            "$ref": "#/components/responses/ArchitectureList"
          }
        }
      }
    },
    "/arch/{arch_id}": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchitecture",
        "summary": "Get architecture metadata",
        "description": "Returns metadata for one registered architecture. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/Architecture"
          }
        }
      }
    },
    "/singer": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getSingerList",
        "summary": "List all singers",
        "description": "Returns singers from every registered architecture. Display language is specified via the Accept-Language request header.",
        "responses": {
          "200": {
            "$ref": "#/components/responses/SingerList"
          }
        }
      }
    },
    "/arch/{arch_id}/singer": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchSingerList",
        "summary": "List singers for an architecture",
        "description": "Returns singers from one registered architecture. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/ArchSingerList"
          }
        }
      }
    },
    "/arch/{arch_id}/singer/{singer_id}": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchSinger",
        "summary": "Get singer metadata",
        "description": "Returns one singer from one registered architecture. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          },
          {
            "$ref": "#/components/parameters/SingerID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/Singer"
          }
        }
      }
    },
    "/arch/{arch_id}/singer/{singer_id}/avatar": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchSingerAvatar",
        "summary": "Get singer avatar URL",
        "description": "Returns the avatar URL for one singer. Display language is specified via the Accept-Language request header. Acceptable formats and schemes can be constrained through the acceptable_formats and acceptable_schemes query parameters.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          },
          {
            "$ref": "#/components/parameters/SingerID"
          },
          {
            "$ref": "#/components/parameters/AcceptableFormats"
          },
          {
            "$ref": "#/components/parameters/AcceptableSchemes"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/SingerAvatar"
          }
        }
      }
    },
    "/arch/{arch_id}/singer/{singer_id}/background": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchSingerBackground",
        "summary": "Get singer background URL",
        "description": "Returns the background image URL for one singer. Display language is specified via the Accept-Language request header. Acceptable formats and schemes can be constrained through the acceptable_formats and acceptable_schemes query parameters.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          },
          {
            "$ref": "#/components/parameters/SingerID"
          },
          {
            "$ref": "#/components/parameters/AcceptableFormats"
          },
          {
            "$ref": "#/components/parameters/AcceptableSchemes"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/SingerBackground"
          }
        }
      }
    },
    "/arch/{arch_id}/singer/{singer_id}/demo-audio": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchSingerDemoAudioList",
        "summary": "List singer demo audio",
        "description": "Returns demo audio entries for one singer. Display language is specified via the Accept-Language request header. Acceptable formats and schemes can be constrained through the acceptable_formats and acceptable_schemes query parameters.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          },
          {
            "$ref": "#/components/parameters/SingerID"
          },
          {
            "$ref": "#/components/parameters/AcceptableFormats"
          },
          {
            "$ref": "#/components/parameters/AcceptableSchemes"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/SingerDemoAudioList"
          }
        }
      }
    },
    "/arch/{arch_id}/group": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchGroupList",
        "summary": "List groups for an architecture",
        "description": "Returns groups from one registered architecture. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/GroupList"
          }
        }
      }
    },
    "/arch/{arch_id}/group/{group_id}": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchGroup",
        "summary": "Get group metadata",
        "description": "Returns one group from one registered architecture. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          },
          {
            "$ref": "#/components/parameters/GroupID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/Group"
          }
        }
      }
    },
    "/arch/{arch_id}/group/{group_id}/singer": {
      "get": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "getArchGroupSingerList",
        "summary": "List singers for a group",
        "description": "Returns singers from one registered group. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ArchID"
          },
          {
            "$ref": "#/components/parameters/GroupID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/GroupSingerList"
          }
        }
      }
    },
    "/synth/env-tag": {
      "post": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "postSynthEnvTag",
        "summary": "Get an environment tag",
        "description": "Returns the environment tag for a synthesis context.",
        "requestBody": {
          "$ref": "#/components/requestBodies/EnvTagRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/EnvTagResult"
          }
        }
      }
    },
    "/synth/pronunciation": {
      "post": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "postSynthPronunciation",
        "summary": "Generate pronunciations",
        "description": "Generates pronunciation candidates for lyric input. Conversion capability is described by pronunciationMode for the selected singing language in the singer's group.",
        "requestBody": {
          "$ref": "#/components/requestBodies/PronunciationRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/PronunciationComplete"
          }
        }
      }
    },
    "/synth/phoneme": {
      "post": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "postSynthPhoneme",
        "summary": "Generate phonemes",
        "description": "Generates phoneme notes from pronunciation notes. Token conversion and onset-labeling capabilities are described by phonemeMode for the selected singing language in the singer's group.",
        "requestBody": {
          "$ref": "#/components/requestBodies/PhonemeRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/PhonemeComplete"
          }
        }
      }
    },
    "/synth/duration": {
      "post": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "postSynthDuration",
        "summary": "Generate phoneme timing",
        "description": "Generates phoneme timing for duration notes when the selected group has durationMode=full. Clients supply phoneme starts directly for later stages when durationMode=skip.",
        "requestBody": {
          "$ref": "#/components/requestBodies/DurationRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/DurationResult"
          }
        }
      }
    },
    "/synth/parameter": {
      "post": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "postSynthParameter",
        "summary": "Generate parameter values",
        "description": "Generates parameter values explicitly requested through retake. Consult the selected group's parameterPipeline for supported inputs, prediction dependencies, and full-piece or subrange retake capabilities. Parameters supplied without retake are conditions.",
        "requestBody": {
          "$ref": "#/components/requestBodies/ParameterRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/ParameterResult"
          }
        }
      }
    },
    "/synth/audio": {
      "post": {
        "tags": [
          "Synthesis"
        ],
        "operationId": "postSynthAudio",
        "summary": "Generate audio",
        "description": "Generates encoded audio from prepared phoneme starts and parameter curves. The selected group's audioDependencies describes required parameter inputs. Formats and URL schemes can be constrained through acceptableFormats and acceptableSchemes.",
        "requestBody": {
          "$ref": "#/components/requestBodies/AudioRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/AudioResult"
          }
        }
      }
    },
    "/extractor": {
      "get": {
        "tags": [
          "Extraction"
        ],
        "operationId": "getExtractorList",
        "summary": "List extractors",
        "description": "Returns the available note, tempo, pitch, and separation extractors. Display language is specified via the Accept-Language request header.",
        "responses": {
          "200": {
            "$ref": "#/components/responses/ExtractorList"
          }
        }
      }
    },
    "/extractor/note/{extractor_id}": {
      "get": {
        "tags": [
          "Extraction"
        ],
        "operationId": "getNoteExtractor",
        "summary": "Get a note extractor",
        "description": "Returns metadata for one note extractor. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ExtractorID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/Extractor"
          }
        }
      }
    },
    "/extractor/tempo/{extractor_id}": {
      "get": {
        "tags": [
          "Extraction"
        ],
        "operationId": "getTempoExtractor",
        "summary": "Get a tempo extractor",
        "description": "Returns metadata for one tempo extractor. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ExtractorID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/Extractor"
          }
        }
      }
    },
    "/extractor/pitch/{extractor_id}": {
      "get": {
        "tags": [
          "Extraction"
        ],
        "operationId": "getPitchExtractor",
        "summary": "Get a pitch extractor",
        "description": "Returns metadata for one pitch extractor. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ExtractorID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/Extractor"
          }
        }
      }
    },
    "/extractor/separation/{extractor_id}": {
      "get": {
        "tags": [
          "Extraction"
        ],
        "operationId": "getSeparationExtractor",
        "summary": "Get a separation extractor",
        "description": "Returns metadata for one separation extractor. Display language is specified via the Accept-Language request header.",
        "parameters": [
          {
            "$ref": "#/components/parameters/ExtractorID"
          }
        ],
        "responses": {
          "200": {
            "$ref": "#/components/responses/SeparationExtractor"
          }
        }
      }
    },
    "/extract/note": {
      "post": {
        "tags": [
          "Extraction"
        ],
        "operationId": "postExtractNote",
        "summary": "Extract notes from audio",
        "description": "Extracts note timing and pitch from audio.",
        "requestBody": {
          "$ref": "#/components/requestBodies/NoteExtractionRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/NoteExtractionResult"
          }
        }
      }
    },
    "/extract/tempo": {
      "post": {
        "tags": [
          "Extraction"
        ],
        "operationId": "postExtractTempo",
        "summary": "Extract tempo from audio",
        "description": "Extracts an ordered list of beats and downbeat indicators from audio.",
        "requestBody": {
          "$ref": "#/components/requestBodies/TempoExtractionRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/TempoExtractionResult"
          }
        }
      }
    },
    "/extract/pitch": {
      "post": {
        "tags": [
          "Extraction"
        ],
        "operationId": "postExtractPitch",
        "summary": "Extract pitch from audio",
        "description": "Extracts a segmented pitch curve from audio.",
        "requestBody": {
          "$ref": "#/components/requestBodies/PitchExtractionRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/PitchExtractionResult"
          }
        }
      }
    },
    "/extract/separation": {
      "post": {
        "tags": [
          "Extraction"
        ],
        "operationId": "postExtractSeparation",
        "summary": "Separate audio",
        "description": "Separates audio into its constituent tracks. Output formats and URL schemes are constrained by acceptableFormats and acceptableSchemes in the request body. The server currently returns only audio/wav data URLs, preserving the input sample rate without resampling; the input must use a sample rate accepted by the extractor. A nonempty preference list that excludes the supported output format or scheme is rejected with HTTP 422. Display language is specified via the Accept-Language request header.",
        "requestBody": {
          "$ref": "#/components/requestBodies/SeparationExtractionRequest"
        },
        "responses": {
          "200": {
            "$ref": "#/components/responses/SeparationExtractionResult"
          }
        }
      }
    }
  },
  "components": {
    "parameters": {
      "ArchID": {
        "name": "arch_id",
        "in": "path",
        "required": true,
        "schema": {
          "type": "string"
        },
        "description": "Architecture identifier."
      },
      "SingerID": {
        "name": "singer_id",
        "in": "path",
        "required": true,
        "schema": {
          "type": "string"
        },
        "description": "Singer reference understood by the selected architecture. DiffSinger uses package-id=version:singer/contribution-id with the exact package version. Preserve the returned reference and encode it once as a URI path component when used in a path."
      },
      "GroupID": {
        "name": "group_id",
        "in": "path",
        "required": true,
        "schema": {
          "type": "string"
        },
        "description": "Group identifier within the selected architecture."
      },
      "ExtractorID": {
        "name": "extractor_id",
        "in": "path",
        "required": true,
        "schema": {
          "type": "string"
        },
        "description": "Extractor identifier within the selected extraction capability."
      },
      "AcceptableFormats": {
        "name": "acceptableFormats",
        "in": "query",
        "required": false,
        "description": "Comma-separated MIME types the client accepts, in priority order. If omitted or empty, the server may return any URL.",
        "style": "form",
        "explode": false,
        "allowEmptyValue": true,
        "schema": {
          "type": "array",
          "items": {
            "type": "string",
            "format": "media-type"
          },
          "description": "MIME types the client expects to accept, in priority order. An empty list means the server may return any URL."
        }
      },
      "AcceptableSchemes": {
        "name": "acceptableSchemes",
        "in": "query",
        "required": false,
        "description": "Comma-separated URL schemes the client accepts, in priority order. If omitted or empty, the server may return any URL.",
        "style": "form",
        "explode": false,
        "allowEmptyValue": true,
        "schema": {
          "type": "array",
          "items": {
            "type": "string"
          },
          "description": "URL schemes the client expects to accept, in priority order. An empty list means the server may return any URL."
        }
      }
    },
    "requestBodies": {
      "EnvTagRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/EnvTagRequest"
            }
          }
        }
      },
      "PronunciationRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/PronunciationRequest"
            }
          }
        }
      },
      "PhonemeRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/PhonemeRequest"
            }
          }
        }
      },
      "DurationRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/DurationRequest"
            }
          }
        }
      },
      "ParameterRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/ParameterRequest"
            }
          }
        }
      },
      "AudioRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/AudioRequest"
            }
          }
        }
      },
      "NoteExtractionRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/NoteExtractionRequest"
            }
          }
        }
      },
      "TempoExtractionRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/TempoExtractionRequest"
            }
          }
        }
      },
      "PitchExtractionRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/PitchExtractionRequest"
            }
          }
        }
      },
      "SeparationExtractionRequest": {
        "required": true,
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SeparationExtractionRequest"
            }
          }
        }
      }
    },
    "responses": {
      "ArchitectureList": {
        "description": "Architecture metadata from every registered architecture.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/ArchitectureMetadataList"
            }
          }
        }
      },
      "Architecture": {
        "description": "Architecture metadata.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/ArchitectureMetadata"
            }
          }
        }
      },
      "SingerList": {
        "description": "Singer metadata from every registered architecture.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerInfoList"
            }
          }
        }
      },
      "ArchSingerList": {
        "description": "Singer metadata from one registered architecture.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerInfoList"
            }
          }
        }
      },
      "Singer": {
        "description": "Singer metadata.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerInfo"
            }
          }
        }
      },
      "SingerAvatar": {
        "description": "Singer avatar URL.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerAvatarResponse"
            }
          }
        }
      },
      "SingerBackground": {
        "description": "Singer background URL.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerBackgroundResponse"
            }
          }
        }
      },
      "SingerDemoAudioList": {
        "description": "Singer demo audio entries.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerDemoAudioList"
            }
          }
        }
      },
      "GroupList": {
        "description": "Group metadata from one registered architecture.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/GroupMetadataList"
            }
          }
        }
      },
      "Group": {
        "description": "Group metadata.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/GroupMetadata"
            }
          }
        }
      },
      "GroupSingerList": {
        "description": "Singer metadata from one registered group.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SingerInfoList"
            }
          }
        }
      },
      "EnvTagResult": {
        "description": "Environment tag for the supplied synthesis context.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/EnvTagResponse"
            }
          }
        }
      },
      "PronunciationComplete": {
        "description": "Pronunciation generation result.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/PronunciationResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/PronunciationResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "PhonemeComplete": {
        "description": "Phoneme generation result.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/PhonemeResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/PhonemeResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "DurationResult": {
        "description": "Duration generation result.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/DurationResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/DurationResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "ParameterResult": {
        "description": "Parameter generation result.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/ParameterResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/ParameterResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "AudioResult": {
        "description": "Audio generation result.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/AudioResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/AudioResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "ExtractorList": {
        "description": "Available extractors grouped by capability.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/ExtractorList"
            }
          }
        }
      },
      "Extractor": {
        "description": "Extractor metadata.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/ExtractorInfo"
            }
          }
        }
      },
      "SeparationExtractor": {
        "description": "Separation extractor metadata, including the available output tracks.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SeparationExtractorInfo"
            }
          }
        }
      },
      "NoteExtractionResult": {
        "description": "Extracted notes.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/NoteExtractionResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/NoteExtractionResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "TempoExtractionResult": {
        "description": "Extracted beats and downbeat indicators.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/TempoExtractionResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/TempoExtractionResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "PitchExtractionResult": {
        "description": "Extracted segmented pitch curve.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/PitchExtractionResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/PitchExtractionResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      },
      "SeparationExtractionResult": {
        "description": "Separated audio tracks.",
        "content": {
          "application/json": {
            "schema": {
              "$ref": "#/components/schemas/SeparationExtractionResponse"
            }
          },
          "application/x-ndjson": {
            "itemSchema": {
              "oneOf": [
                {
                  "$ref": "#/components/schemas/QueueEvent"
                },
                {
                  "$ref": "#/components/schemas/ProgressEvent"
                },
                {
                  "type": "object",
                  "additionalProperties": false,
                  "required": [
                    "event",
                    "data"
                  ],
                  "properties": {
                    "event": {
                      "const": "result"
                    },
                    "data": {
                      "$ref": "#/components/schemas/SeparationExtractionResponse"
                    }
                  }
                },
                {
                  "$ref": "#/components/schemas/ErrorEvent"
                }
              ]
            }
          }
        }
      }
    },
    "schemas": {
      "QueueEvent": {
        "type": "object",
        "description": "Indicates initial resource admission. Heartbeats may repeat this event until processing begins. Subsequent resource waits retain the progress state. Disconnecting in this state does not discard unfinished processing progress. Any event may follow.",
        "additionalProperties": false,
        "required": [
          "event"
        ],
        "properties": {
          "event": {
            "const": "queue"
          }
        }
      },
      "ProgressEvent": {
        "type": "object",
        "description": "Indicates that the task is processing. Absent data is equivalent to {\"value\": 0}. Values are monotonic and heartbeats may repeat the current value. A value of 1 indicates task body completion. Result construction may still fail. Only progress, result, or error may follow.",
        "additionalProperties": false,
        "required": [
          "event"
        ],
        "properties": {
          "event": {
            "const": "progress"
          },
          "data": {
            "type": "object",
            "additionalProperties": false,
            "required": [
              "value"
            ],
            "properties": {
              "value": {
                "type": "number",
                "minimum": 0,
                "maximum": 1,
                "description": "Normalized progress for client display only. The value has no business meaning."
              }
            },
            "default": {
              "value": 0
            }
          }
        }
      },
      "ErrorEvent": {
        "type": "object",
        "description": "Returns a task failure. The error data format is unspecified. This event is terminal.",
        "additionalProperties": false,
        "required": [
          "event",
          "data"
        ],
        "properties": {
          "event": {
            "const": "error"
          },
          "data": {}
        }
      },
      "ExtractorList": {
        "type": "object",
        "required": [
          "note",
          "tempo",
          "pitch",
          "separation"
        ],
        "properties": {
          "note": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ExtractorInfo"
            }
          },
          "tempo": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ExtractorInfo"
            }
          },
          "pitch": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ExtractorInfo"
            }
          },
          "separation": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/SeparationExtractorInfo"
            }
          }
        }
      },
      "ExtractorInfo": {
        "type": "object",
        "required": [
          "id",
          "name",
          "preferredAudioSampleRate"
        ],
        "properties": {
          "id": {
            "type": "string",
            "description": "Extractor identifier within its extraction capability."
          },
          "name": {
            "type": "string"
          },
          "preferredAudioSampleRate": {
            "type": "integer",
            "exclusiveMinimum": 0,
            "description": "Preferred input audio sample rate in Hz."
          },
          "acceptableFormats": {
            "type": "array",
            "items": {
              "type": "string",
              "format": "media-type"
            },
            "description": "MIME types this extractor can accept, in priority order. If omitted or empty, the extractor can accept any format."
          },
          "acceptableSchemes": {
            "type": "array",
            "items": {
              "type": "string"
            },
            "description": "URL schemes this extractor can accept, in priority order. If omitted or empty, the extractor can accept any scheme."
          },
          "acceptableAudioSampleRates": {
            "type": "array",
            "items": {
              "type": "integer",
              "exclusiveMinimum": 0
            },
            "description": "Input audio sample rates in Hz that this extractor can accept, in priority order. If omitted or empty, the extractor can accept any sample rate."
          }
        }
      },
      "SeparationExtractorInfo": {
        "allOf": [
          {
            "$ref": "#/components/schemas/ExtractorInfo"
          },
          {
            "type": "object",
            "required": [
              "tracks"
            ],
            "properties": {
              "tracks": {
                "type": "array",
                "items": {
                  "$ref": "#/components/schemas/SeparationTrackInfo"
                },
                "description": "Available output tracks in the same order as the separation result."
              }
            }
          }
        ]
      },
      "SeparationTrackInfo": {
        "type": "object",
        "additionalProperties": false,
        "required": [
          "name"
        ],
        "properties": {
          "name": {
            "type": "string",
            "description": "Stable, non-localized track name matching name in the separation result, for example drums, bass, other or vocals."
          }
        }
      },
      "NoteExtractionRequest": {
        "$ref": "#/components/schemas/ExtractionRequest"
      },
      "TempoExtractionRequest": {
        "$ref": "#/components/schemas/ExtractionRequest"
      },
      "PitchExtractionRequest": {
        "$ref": "#/components/schemas/ExtractionRequest"
      },
      "SeparationExtractionRequest": {
        "allOf": [
          {
            "$ref": "#/components/schemas/ExtractionRequest"
          },
          {
            "type": "object",
            "properties": {
              "acceptableFormats": {
                "type": "array",
                "items": {
                  "type": "string",
                  "format": "media-type"
                },
                "description": "MIME types the client accepts for every separated track, in priority order. If omitted or empty, the server may return any supported format. MIME types are compared case-insensitively and must be type/subtype without parameters or wildcards. The server currently supports only audio/wav."
              },
              "acceptableSchemes": {
                "type": "array",
                "items": {
                  "type": "string"
                },
                "description": "URL schemes the client accepts for every separated track, in priority order. If omitted or empty, the server may return any supported scheme. Schemes are compared case-insensitively. The server currently supports only data."
              }
            }
          }
        ]
      },
      "ExtractionRequest": {
        "type": "object",
        "required": [
          "extractor",
          "input"
        ],
        "properties": {
          "extractor": {
            "type": "string",
            "description": "Identifier of an extractor for the requested capability."
          },
          "input": {
            "$ref": "#/components/schemas/ExtractionInput"
          }
        }
      },
      "ExtractionInput": {
        "type": "object",
        "required": [
          "audioUrl"
        ],
        "properties": {
          "audioUrl": {
            "type": "string",
            "format": "uri",
            "description": "URL of the audio to analyze."
          }
        }
      },
      "NoteExtractionResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/NoteExtractionOutput"
          }
        }
      },
      "NoteExtractionOutput": {
        "type": "object",
        "required": [
          "notes"
        ],
        "properties": {
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ExtractedNote"
            }
          }
        }
      },
      "ExtractedNote": {
        "type": "object",
        "required": [
          "position",
          "cent"
        ],
        "properties": {
          "position": {
            "$ref": "#/components/schemas/NotePosition"
          },
          "cent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 12800,
            "description": "Extracted note pitch in cents."
          }
        }
      },
      "TempoExtractionResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/TempoExtractionOutput"
          }
        }
      },
      "TempoExtractionOutput": {
        "type": "object",
        "required": [
          "beats"
        ],
        "properties": {
          "beats": {
            "type": "array",
            "description": "Extracted beats in chronological order. The array can be empty when no beats are detected.",
            "items": {
              "$ref": "#/components/schemas/ExtractedBeat"
            }
          }
        }
      },
      "ExtractedBeat": {
        "type": "object",
        "required": [
          "position",
          "downbeat"
        ],
        "properties": {
          "position": {
            "type": "number",
            "minimum": 0,
            "description": "Absolute time in seconds from the start of the input audio to this beat."
          },
          "downbeat": {
            "type": "boolean",
            "description": "Whether this beat is a downbeat (the first beat of a measure)."
          }
        }
      },
      "PitchExtractionResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/PitchExtractionOutput"
          }
        }
      },
      "PitchExtractionOutput": {
        "type": "object",
        "required": [
          "segments",
          "sampleRate"
        ],
        "properties": {
          "segments": {
            "type": "array",
            "description": "Voiced pitch segments. The array can be empty when no pitch is detected.",
            "items": {
              "$ref": "#/components/schemas/PitchExtractionSegment"
            }
          },
          "sampleRate": {
            "type": "number",
            "exclusiveMinimum": 0,
            "description": "Number of pitch samples per second."
          }
        }
      },
      "SeparationExtractionResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/SeparationExtractionOutput"
          }
        }
      },
      "SeparationExtractionOutput": {
        "type": "object",
        "required": [
          "tracks"
        ],
        "properties": {
          "tracks": {
            "type": "array",
            "description": "Separated audio tracks. The array can be empty when no tracks are produced.",
            "items": {
              "$ref": "#/components/schemas/SeparationTrack"
            }
          }
        }
      },
      "SeparationTrack": {
        "type": "object",
        "required": [
          "name",
          "audioUrl"
        ],
        "properties": {
          "name": {
            "type": "string"
          },
          "audioUrl": {
            "type": "string",
            "description": "Encoded audio URL."
          }
        }
      },
      "PitchExtractionSegment": {
        "type": "object",
        "required": [
          "gap",
          "pitch"
        ],
        "properties": {
          "gap": {
            "type": "integer",
            "minimum": 0,
            "description": "Number of missing pitch frames before this segment. For the first segment, frames are counted from the start of the audio. For later segments, the start frame is the previous segment start plus its pitch value count plus this gap."
          },
          "pitch": {
            "type": "array",
            "minItems": 1,
            "description": "Pitch values in cents.",
            "items": {
              "type": "number",
              "minimum": 0,
              "maximum": 12800
            }
          }
        }
      },
      "ApplicationInfoResponse": {
        "type": "object",
        "required": [
          "dssp"
        ],
        "properties": {
          "dssp": {
            "$ref": "#/components/schemas/ApplicationInfo"
          }
        }
      },
      "ApplicationInfo": {
        "type": "object",
        "required": [
          "apiVersion"
        ],
        "properties": {
          "apiVersion": {
            "type": "object",
            "required": [
              "major",
              "minor"
            ],
            "properties": {
              "major": {
                "type": "number",
                "description": "Major API version number."
              },
              "minor": {
                "type": "number",
                "description": "Minor API version number."
              }
            },
            "description": "API version represented by major and minor numbers."
          }
        }
      },
      "ArchitectureMetadataList": {
        "type": "array",
        "items": {
          "$ref": "#/components/schemas/ArchitectureMetadata"
        }
      },
      "ArchitectureMetadata": {
        "type": "object",
        "required": [
          "id",
          "name",
          "parameters",
          "defaultArchExtra"
        ],
        "properties": {
          "id": {
            "type": "string",
            "description": "Architecture identifier."
          },
          "name": {
            "type": "string"
          },
          "parameters": {
            "type": "object",
            "description": "Parameter definitions keyed by parameter ID.",
            "additionalProperties": {
              "type": "object",
              "required": [
                "name",
                "defaultValue",
                "showBaseline",
                "baselineValue",
                "fillMode",
                "fallbackOnDefaultValue",
                "displayValueMappingExpression",
                "displayValueInverseMappingExpression",
                "displayValuePrefix",
                "displayValueSuffix",
                "displayValueDecimalPlaces"
              ],
              "properties": {
                "name": {
                  "type": "string",
                  "description": "Human-readable parameter name."
                },
                "defaultValue": {
                  "type": "number",
                  "minimum": 0,
                  "maximum": 1,
                  "description": "Default parameter value used when a parameter curve has no sample value."
                },
                "showBaseline": {
                  "type": "boolean",
                  "description": "Whether clients should display the parameter baseline."
                },
                "baselineValue": {
                  "type": "number",
                  "minimum": 0,
                  "maximum": 1,
                  "description": "Baseline parameter value."
                },
                "fillMode": {
                  "type": "string",
                  "enum": [
                    "no",
                    "top",
                    "bottom",
                    "baseline"
                  ],
                  "description": "How clients should fill the area associated with the parameter curve."
                },
                "fallbackOnDefaultValue": {
                  "type": "boolean",
                  "description": "Whether clients should use the default value when drawing curve segments that have no sample values."
                },
                "displayValueMappingExpression": {"$ref":"#/components/schemas/AST"},
                "displayValueInverseMappingExpression": {"$ref":"#/components/schemas/AST"},
                "displayValuePrefix": {
                  "type": "string",
                  "description": "Prefix to include when displaying the mapped value as text."
                },
                "displayValueSuffix": {
                  "type": "string",
                  "description": "Suffix to include when displaying the mapped value as text."
                },
                "displayValueDecimalPlaces": {
                  "type": "integer",
                  "minimum": 0,
                  "description": "Number of decimal places to use when displaying the mapped value as text."
                }
              }
            }
          },
          "defaultArchExtra": {
            "description": "Default architecture-specific JSON payload for context.archExtra. May be any JSON value, including null. The selected architecture defines its meaning. These defaults apply to the shared request context for all groups of this architecture and are separate from each singer's defaultExtra.",
            "examples": [
              {
                "steps": {
                  "pitch": 20,
                  "variance": 20,
                  "acoustic": 20
                },
                "depth": 1
              }
            ]
          }
        }
      },
      "GroupMetadataList": {
        "type": "array",
        "items": {
          "$ref": "#/components/schemas/GroupMetadata"
        }
      },
      "GroupMetadata": {
        "type": "object",
        "description": "A shared complete synthesis pipeline and its effective capabilities. All singers in a group have the same supported singing languages, language conversion bindings, and pipeline capabilities. Singer-specific voice selections and default languages may differ. Metadata discovery does not require loading inference models.",
        "required": [
          "id",
          "arch",
          "languages",
          "durationMode",
          "parameterPipeline",
          "audioDependencies",
          "mixable",
          "archSpecificInfo"
        ],
        "properties": {
          "id": {
            "type": "string",
            "description": "Group identifier within the selected architecture."
          },
          "arch": {
            "type": "string",
            "description": "Architecture identifier."
          },
          "languages": {
            "type": "object",
            "additionalProperties": {
              "$ref": "#/components/schemas/GroupLanguageInfo"
            },
            "description": "Supported singing language handles mapped to their localized metadata and individual conversion capabilities. An empty object means no singing language is available. Different languages may have different conversion capabilities; there is no aggregate conversion mode on the group."
          },
          "durationMode": {
            "$ref": "#/components/schemas/DurationMode"
          },
          "parameterPipeline": {
            "type": "object",
            "additionalProperties": {
              "$ref": "#/components/schemas/ArchitectureParameterMetadata"
            },
            "description": "Supported parameter inputs and prediction capabilities. Each key must be a parameter ID defined by the selected architecture. A parameter omitted from this map is unsupported by this group. Direct and indirect parameters can both be supplied as input curves; only indirect parameters can be predicted. Dependencies must refer to entries in this map and must be acyclic."
          },
          "audioDependencies": {
            "type": "array",
            "items": {
              "type": "string"
            },
            "description": "Parameter IDs consumed by the audio synthesis stage. Every ID must be present in parameterPipeline. This list describes prepared input requirements; the audio endpoint does not implicitly predict these parameters."
          },
          "mixable": {
            "type": "boolean",
            "description": "Whether multiple singer sources within this group may be mixed. All sources in a multi-singer context must belong to the same architecture and the same group, and that group must have mixable=true. A value of false prohibits mixing within the group; it does not restrict single-source use. This flag does not permit mixing across groups."
          },
          "archSpecificInfo": {
            "description": "Architecture-specific information about the shared synthesis pipeline of this group. May be any JSON value, including null; its meaning is defined by the selected architecture. Common language, duration, and parameter capabilities are described by the dedicated group fields."
          }
        },
        "examples": [
          {
            "id": "example-group",
            "arch": "diffsinger",
            "languages": {
              "cmn": {
                "name": "Mandarin Chinese",
                "defaultLyric": "",
                "pronunciationMode": "full",
                "phonemeMode": "full"
              },
              "eng": {
                "name": "English",
                "defaultLyric": "",
                "pronunciationMode": "full",
                "phonemeMode": "token_only"
              }
            },
            "durationMode": "full",
            "parameterPipeline": {
              "pitch": {
                "type": "indirect",
                "dependsOn": [
                  "expressiveness"
                ],
                "retakeMode": "range"
              },
              "expressiveness": {
                "type": "direct"
              },
              "energy": {
                "type": "indirect",
                "dependsOn": [
                  "pitch"
                ],
                "retakeMode": "range"
              }
            },
            "audioDependencies": [
              "pitch",
              "energy"
            ],
            "mixable": false,
            "archSpecificInfo": {}
          }
        ]
      },
      "GroupLanguageInfo": {
        "type": "object",
        "description": "Language metadata and effective conversion capabilities shared by all singers in the group. The containing map key is the singing language handle accepted by the selected architecture; it is independent of the display locale selected by Accept-Language.",
        "required": [
          "name",
          "defaultLyric",
          "pronunciationMode",
          "phonemeMode"
        ],
        "properties": {
          "name": {
            "type": "string",
            "description": "Localized display name of the singing language."
          },
          "defaultLyric": {
            "type": "string",
            "description": "Default placeholder lyric provided by the language metadata. An empty string means no default lyric is available."
          },
          "pronunciationMode": {
            "$ref": "#/components/schemas/PronunciationMode"
          },
          "phonemeMode": {
            "$ref": "#/components/schemas/PhonemeMode"
          }
        }
      },
      "PronunciationMode": {
        "type": "string",
        "enum": [
          "full",
          "skip"
        ],
        "description": "Per-language automatic lyric-to-pronunciation capability. full means the service can perform the conversion; skip means clients must provide pronunciation directly when required by subsequent stages. skip does not imply a substitute conversion algorithm."
      },
      "PhonemeMode": {
        "type": "string",
        "enum": [
          "full",
          "token_only",
          "skip"
        ],
        "description": "Per-language automatic pronunciation-to-phoneme capability. full supports phoneme tokens and syllable onset labeling; token_only supports phoneme tokens without an onset-labeling capability; skip means clients must provide phonemes directly when required by subsequent stages."
      },
      "DurationMode": {
        "type": "string",
        "enum": [
          "full",
          "skip"
        ],
        "description": "Phoneme timing prediction capability of the shared pipeline. full means the duration endpoint can predict phoneme starts; skip means clients must supply phoneme starts for subsequent stages. A duration prediction request for a group with skip is unsupported and must not be replaced with fabricated timing."
      },
      "ArchitectureParameterMetadata": {
        "oneOf": [
          {
            "$ref": "#/components/schemas/DirectArchitectureParameterMetadata"
          },
          {
            "$ref": "#/components/schemas/IndirectArchitectureParameterMetadata"
          }
        ]
      },
      "DirectArchitectureParameterMetadata": {
        "type": "object",
        "required": [
          "type"
        ],
        "properties": {
          "type": {
            "const": "direct"
          }
        },
        "description": "A supported parameter supplied as an input curve without model prediction. A nonempty retake range for a direct parameter is invalid. A zero-length retake is a no-op."
      },
      "IndirectArchitectureParameterMetadata": {
        "type": "object",
        "required": [
          "type",
          "dependsOn",
          "retakeMode"
        ],
        "properties": {
          "type": {
            "const": "indirect"
          },
          "dependsOn": {
            "type": "array",
            "items": {
              "type": "string"
            },
            "description": "Parameter IDs used as prediction conditions. Dependencies may be direct or indirect and must be present in the group's parameterPipeline."
          },
          "retakeMode": {
            "$ref": "#/components/schemas/ParameterRetakeMode"
          }
        },
        "description": "A supported parameter that can be predicted and can also be supplied as a conditioning curve. Prediction is requested only when the input parameter includes retake; dependencies supplied without retake are conditions and are not implicitly regenerated."
      },
      "ParameterRetakeMode": {
        "type": "string",
        "enum": [
          "full",
          "range"
        ],
        "description": "Supported scope of parameter recomputation exposed by the service. full accepts nonempty retake ranges only when they cover the complete piece; range accepts both subranges and the complete piece. An unsupported nonempty subrange is rejected with 422 and must not be silently expanded. A zero-length retake is a no-op in either mode. This describes observable service behavior, not whether a backend computes only the requested frames."
      },
      "ArchitectureParameterType": {
        "type": "string",
        "enum": [
          "direct",
          "indirect"
        ]
      },
      "SingerInfoList": {
        "type": "array",
        "items": {
          "$ref": "#/components/schemas/SingerInfo"
        }
      },
      "SingerInfo": {
        "type": "object",
        "required": [
          "id",
          "name",
          "arch",
          "group",
          "defaultLanguage",
          "archSpecificInfo",
          "defaultExtra"
        ],
        "properties": {
          "id": {
            "type": "string",
            "description": "Singer reference understood by the selected architecture. DiffSinger uses package-id=version:singer/contribution-id with the exact package version. Preserve the returned reference and encode it once as a URI path component when used in a path."
          },
          "name": {
            "type": "string"
          },
          "arch": {
            "type": "string",
            "description": "Architecture identifier."
          },
          "group": {
            "type": "string",
            "description": "Group identifier within the selected architecture."
          },
          "defaultLanguage": {
            "type": "string",
            "description": "This singer's preferred singing language handle. A nonempty value must be a key in its group's languages map. It may differ between singers in the same group. An empty string means no default language is declared or the group has no available languages; clients must not infer a preference from map order."
          },
          "archSpecificInfo": {
            "description": "Architecture-specific properties of this individual singer, such as selectable voices. DiffSinger returns a speakers array of objects with id and name string fields. The id identifies a selectable logical speaker and is accepted by extra.speaker. The name contains the display name and currently equals the id. The array is empty if no speaker selection is required. Shared duration and language conversion capabilities belong to GroupMetadata and GroupLanguageInfo, not this payload."
          },
          "defaultExtra": {
            "description": "Default architecture-specific parameters for this singer's extra payload, such as a voice selection. Request-wide synthesis parameters are supplied separately in context.archExtra and are not part of these defaults."
          }
        },
        "description": "An installed singer and its individual metadata. Shared language and pipeline capabilities are obtained from the group identified by arch and group.",
        "examples": [
          {
            "id": "example/singers=1.0:singer/main",
            "name": "Example Singer",
            "arch": "diffsinger",
            "group": "example-group",
            "defaultLanguage": "cmn",
            "archSpecificInfo": {
              "speakers": [
                {
                  "id": "main",
                  "name": "main"
                },
                {
                  "id": "soft",
                  "name": "soft"
                }
              ]
            },
            "defaultExtra": {
              "speaker": "main"
            }
          }
        ]
      },
      "SingerAvatarResponse": {
        "type": "object",
        "required": [
          "avatarUrl"
        ],
        "properties": {
          "avatarUrl": {
            "type": "string"
          }
        }
      },
      "SingerBackgroundResponse": {
        "type": "object",
        "required": [
          "backgroundUrl"
        ],
        "properties": {
          "backgroundUrl": {
            "type": "string"
          }
        }
      },
      "SingerDemoAudioList": {
        "type": "array",
        "items": {
          "$ref": "#/components/schemas/SingerDemoAudio"
        }
      },
      "SingerDemoAudio": {
        "type": "object",
        "required": [
          "name",
          "audioUrl"
        ],
        "properties": {
          "name": {
            "type": "string"
          },
          "audioUrl": {
            "type": "string"
          }
        }
      },
      "EnvTagRequest": {
        "type": "object",
        "required": [
          "context"
        ],
        "properties": {
          "context": {
            "$ref": "#/components/schemas/MultiSingerContext"
          }
        }
      },
      "EnvTagResponse": {
        "type": "object",
        "required": [
          "envTag"
        ],
        "properties": {
          "envTag": {
            "$ref": "#/components/schemas/EnvTag"
          }
        }
      },
      "PronunciationRequest": {
        "type": "object",
        "required": [
          "context",
          "input"
        ],
        "properties": {
          "context": {
            "$ref": "#/components/schemas/SingleSingerContext"
          },
          "input": {
            "$ref": "#/components/schemas/PronunciationInput"
          }
        }
      },
      "PhonemeRequest": {
        "type": "object",
        "required": [
          "context",
          "input"
        ],
        "properties": {
          "context": {
            "$ref": "#/components/schemas/SingleSingerContext"
          },
          "input": {
            "$ref": "#/components/schemas/PhonemeInput"
          }
        }
      },
      "DurationRequest": {
        "type": "object",
        "required": [
          "context",
          "input"
        ],
        "properties": {
          "context": {
            "$ref": "#/components/schemas/MultiSingerContext"
          },
          "input": {
            "$ref": "#/components/schemas/DurationInput"
          }
        }
      },
      "ParameterRequest": {
        "type": "object",
        "required": [
          "context",
          "input"
        ],
        "properties": {
          "context": {
            "$ref": "#/components/schemas/MultiSingerContext"
          },
          "input": {
            "$ref": "#/components/schemas/ParameterInput"
          }
        }
      },
      "AudioRequest": {
        "type": "object",
        "required": [
          "context",
          "input"
        ],
        "properties": {
          "context": {
            "$ref": "#/components/schemas/MultiSingerContext"
          },
          "input": {
            "$ref": "#/components/schemas/AudioInput"
          },
          "acceptableFormats": {
            "type": "array",
            "items": {
              "type": "string",
              "format": "media-type"
            },
            "description": "MIME types the client expects to accept, in priority order. An empty list means the server may return any URL."
          },
          "acceptableSchemes": {
            "type": "array",
            "items": {
              "type": "string"
            },
            "description": "URL schemes the client expects to accept, in priority order. An empty list means the server may return any URL."
          }
        }
      },
      "SingleSingerContext": {
        "type": "object",
        "required": [
          "arch",
          "archExtra",
          "singer"
        ],
        "properties": {
          "arch": {
            "type": "string",
            "description": "Architecture identifier."
          },
          "archExtra": {
            "description": "Architecture-specific synthesis parameters shared by the entire request, including all singer sources. DiffSinger uses an object with steps (an object with optional pitch, variance, and acoustic fields, each a positive signed 64-bit integer with default 20) and depth (a finite number from 0 to 1 with default 1, used only by acoustic inference). Each steps field controls sampling in the corresponding inference stage. Omitting steps, supplying an empty steps object, or supplying an empty archExtra object selects the corresponding defaults. The selected architecture interprets and validates this payload."
          },
          "singer": {
            "$ref": "#/components/schemas/Singer"
          }
        }
      },
      "MultiSingerContext": {
        "type": "object",
        "required": [
          "arch",
          "archExtra",
          "singers"
        ],
        "properties": {
          "arch": {
            "type": "string",
            "description": "Architecture identifier."
          },
          "archExtra": {
            "description": "Architecture-specific synthesis parameters shared by the entire request, including all singer sources. DiffSinger uses an object with steps (an object with optional pitch, variance, and acoustic fields, each a positive signed 64-bit integer with default 20) and depth (a finite number from 0 to 1 with default 1, used only by acoustic inference). Each steps field controls sampling in the corresponding inference stage. Omitting steps, supplying an empty steps object, or supplying an empty archExtra object selects the corresponding defaults. The selected architecture interprets and validates this payload."
          },
          "singers": {
            "type": "array",
            "minItems": 1,
            "items": {
              "$ref": "#/components/schemas/Singer"
            },
            "description": "Ordered singer sources. Multiple sources must belong to the same architecture and group, and that group must allow mixing. Each source has its own singer-specific extra payload; all sources share context.archExtra."
          }
        }
      },
      "Singer": {
        "type": "object",
        "required": [
          "id",
          "extra"
        ],
        "properties": {
          "id": {
            "type": "string",
            "description": "Singer reference understood by the selected architecture. DiffSinger uses package-id=version:singer/contribution-id with the exact package version. Preserve the returned reference and encode it once as a URI path component when used in a path."
          },
          "extra": {
            "description": "Architecture-specific parameters for this singer source, such as a voice selection. SingerInfo.defaultExtra provides the corresponding defaults. Request-wide synthesis parameters are supplied separately in context.archExtra. The selected architecture interprets and validates this payload."
          }
        }
      },
      "EnvTag": {
        "type": "string",
        "description": "Environment tag for a synthesis context."
      },
      "PronunciationInput": {
        "type": "object",
        "required": [
          "notes"
        ],
        "properties": {
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/Lyric"
            }
          }
        }
      },
      "Lyric": {
        "type": "object",
        "required": [
          "lyric",
          "language"
        ],
        "properties": {
          "lyric": {
            "type": "string",
            "description": "This note's own lyric. Every input note supplies its own lyric and language; lyric text does not encode slur relationships or syllable allocation markers."
          },
          "language": {
            "type": "string"
          }
        }
      },
      "PhonemeInput": {
        "type": "object",
        "required": [
          "notes"
        ],
        "properties": {
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/PronunciationNote"
            }
          }
        },
        "description": "Pronunciation input for individual notes. Each item supplies its own pronunciation and language, and may select a syllable range using syllableSliceStart and syllableSliceEnd."
      },
      "PronunciationNote": {
        "type": "object",
        "required": [
          "pronunciation",
          "language"
        ],
        "properties": {
          "pronunciation": {
            "type": "string",
            "description": "This note's own complete pronunciation. Adjacent notes may repeat the same pronunciation without cross-note references. Syllable ranges are expressed by slice fields, not by plus signs or suffixes in the pronunciation text."
          },
          "language": {
            "type": "string"
          },
          "syllableSliceStart": {
            "$ref": "#/components/schemas/SyllableSliceEndpoint",
            "description": "Inclusive start of the syllable range. Omission or null selects position 0. Integer endpoint conversion and clamping are defined by SyllableSliceEndpoint. Positions count syllable groups, not phonemes."
          },
          "syllableSliceEnd": {
            "$ref": "#/components/schemas/SyllableSliceEndpoint",
            "description": "Exclusive end of the syllable range. Omission or null selects M, the number of groups in the complete result. Integer endpoint conversion and clamping are defined by SyllableSliceEndpoint. The selected range is empty when its start is greater than or equal to its end."
          }
        },
        "description": "Slice endpoints refer to syllable groups in this note's complete phoneme result. The first group begins at the start of the sequence and includes phonemes preceding the first onset=true. The second and each subsequent onset=true begin a new group. A nonempty sequence with no onset=true forms one group; an empty sequence has zero groups. All phonemes in the selected groups are returned in their original order with their token and onset values preserved."
      },
      "DurationInput": {
        "type": "object",
        "required": [
          "pieceDuration",
          "notes",
          "mix",
          "mixSampleRate"
        ],
        "properties": {
          "pieceDuration": {
            "type": "number",
            "minimum": 0
          },
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/DurationNote"
            }
          },
          "mix": {
            "$ref": "#/components/schemas/Mix"
          },
          "mixSampleRate": {
            "type": "number",
            "exclusiveMinimum": 0
          }
        }
      },
      "DurationNote": {
        "type": "object",
        "required": [
          "position",
          "cent",
          "pronunciation",
          "language",
          "phonemes",
          "kind"
        ],
        "properties": {
          "position": {
            "$ref": "#/components/schemas/NotePosition"
          },
          "cent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 12800
          },
          "pronunciation": {
            "type": "string",
            "description": "This note's complete pronunciation. Slur notes also supply their own pronunciation and do not use a hyphen placeholder."
          },
          "language": {
            "type": "string"
          },
          "phonemes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/DurationInputPhoneme"
            }
          },
          "kind": {
            "$ref": "#/components/schemas/NoteKind"
          }
        },
        "description": "Each item represents one note and supplies its own pronunciation and language. phonemes contains the phonemes allocated to this note; kind identifies a normal pronunciation or a slur relationship."
      },
      "NotePosition": {
        "type": "object",
        "required": [
          "gap",
          "duration"
        ],
        "properties": {
          "gap": {
            "type": "number",
            "minimum": 0
          },
          "duration": {
            "type": "number",
            "minimum": 0
          }
        }
      },
      "DurationInputPhoneme": {
        "type": "object",
        "required": [
          "token",
          "onset",
          "language"
        ],
        "properties": {
          "token": {
            "type": "string"
          },
          "onset": {
            "type": "boolean"
          },
          "language": {
            "type": "string"
          }
        }
      },
      "ParameterInput": {
        "type": "object",
        "required": [
          "pieceDuration",
          "notes",
          "mix",
          "mixSampleRate",
          "parameters"
        ],
        "properties": {
          "pieceDuration": {
            "type": "number",
            "minimum": 0
          },
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ParameterNote"
            }
          },
          "mix": {
            "$ref": "#/components/schemas/Mix"
          },
          "mixSampleRate": {
            "type": "number",
            "exclusiveMinimum": 0
          },
          "parameters": {
            "$ref": "#/components/schemas/ParameterMap"
          }
        }
      },
      "ParameterNote": {
        "type": "object",
        "required": [
          "position",
          "cent",
          "pronunciation",
          "language",
          "phonemes",
          "kind"
        ],
        "properties": {
          "position": {
            "$ref": "#/components/schemas/NotePosition"
          },
          "cent": {
            "type": "integer",
            "minimum": 0,
            "maximum": 12800
          },
          "pronunciation": {
            "type": "string",
            "description": "This note's complete pronunciation. Slur notes also supply their own pronunciation and do not use a hyphen placeholder."
          },
          "language": {
            "type": "string"
          },
          "phonemes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ParameterInputPhoneme"
            }
          },
          "kind": {
            "$ref": "#/components/schemas/NoteKind"
          }
        },
        "description": "Each item represents one note and supplies its own pronunciation and language. phonemes contains the phonemes allocated to this note; kind identifies a normal pronunciation or a slur relationship."
      },
      "ParameterInputPhoneme": {
        "type": "object",
        "required": [
          "token",
          "onset",
          "language",
          "start"
        ],
        "properties": {
          "token": {
            "type": "string"
          },
          "onset": {
            "type": "boolean"
          },
          "language": {
            "type": "string"
          },
          "start": {
            "type": "number"
          }
        }
      },
      "AudioInput": {
        "type": "object",
        "required": [
          "pieceDuration",
          "notes",
          "mix",
          "mixSampleRate",
          "parameters"
        ],
        "properties": {
          "pieceDuration": {
            "type": "number",
            "minimum": 0
          },
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/ParameterNote"
            }
          },
          "mix": {
            "$ref": "#/components/schemas/Mix"
          },
          "mixSampleRate": {
            "type": "number",
            "exclusiveMinimum": 0
          },
          "parameters": {
            "$ref": "#/components/schemas/AudioParameterMap"
          }
        }
      },
      "Mix": {
        "type": "array",
        "description": "Per-frame singer mix data. Each value must be in [0, 1]. Each row must contain exactly singers.length - 1 values, and the sum of values in each row must be less than or equal to 1.",
        "items": {
          "type": "array",
          "items": {
            "type": "number",
            "minimum": 0,
            "maximum": 1
          }
        }
      },
      "ParameterMap": {
        "type": "object",
        "description": "Open string-keyed parameter map. Supported keys and semantic requirements are defined by the selected architecture.",
        "additionalProperties": {
          "$ref": "#/components/schemas/Parameter"
        }
      },
      "AudioParameterMap": {
        "type": "object",
        "description": "Open string-keyed audio parameter map. Supported keys and semantic requirements are defined by the selected architecture.",
        "additionalProperties": {
          "$ref": "#/components/schemas/AudioParameter"
        }
      },
      "Parameter": {
        "type": "object",
        "required": [
          "values",
          "sampleRate"
        ],
        "properties": {
          "values": {
            "type": "array",
            "items": {
              "type": "number"
            }
          },
          "sampleRate": {
            "type": "number",
            "exclusiveMinimum": 0
          },
          "retake": {
            "$ref": "#/components/schemas/ParameterRetake",
            "description": "Presence requests recomputation of this parameter; omission supplies only a condition curve. A nonempty range is valid only for an indirect parameter and must satisfy its group's retakeMode. A zero-length range returns the existing effective curve without recomputation."
          }
        }
      },
      "AudioParameter": {
        "type": "object",
        "required": [
          "sampleRate"
        ],
        "properties": {
          "values": {
            "type": "array",
            "items": {
              "type": "number"
            }
          },
          "sampleRate": {
            "type": "number",
            "exclusiveMinimum": 0
          }
        }
      },
      "ParameterRetake": {
        "type": "object",
        "required": [
          "position",
          "length"
        ],
        "properties": {
          "position": {
            "type": "integer",
            "minimum": 0,
            "description": "Inclusive start frame on this parameter's input sampling grid."
          },
          "length": {
            "type": "integer",
            "minimum": 0,
            "description": "Number of frames to recompute on this parameter's input sampling grid; zero means no recomputation."
          }
        },
        "description": "Half-open recomputation range in frames of this parameter's input sampleRate: [position/sampleRate, (position+length)/sampleRate) seconds. The range must stay within pieceDuration, allowing the normal final-frame rounding for a full-piece range. Group parameter metadata determines whether subranges are supported; full mode requires every nonempty range to cover the complete piece. A zero-length range changes nothing."
      },
      "PronunciationResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/PronunciationOutput"
          }
        }
      },
      "PronunciationOutput": {
        "type": "object",
        "required": [
          "notes"
        ],
        "properties": {
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/Pronunciation"
            }
          }
        }
      },
      "Pronunciation": {
        "type": "object",
        "required": [
          "pronunciation",
          "candidates"
        ],
        "properties": {
          "pronunciation": {
            "type": "string",
            "description": "The complete pronunciation selected for this note, without appended slur, syllable quota, or allocation markers."
          },
          "candidates": {
            "type": "array",
            "items": {
              "type": "string"
            },
            "description": "Complete pronunciation candidates for this note, without appended slur or syllable allocation markers."
          }
        }
      },
      "PhonemeResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/PhonemeOutput"
          }
        }
      },
      "PhonemeOutput": {
        "type": "object",
        "required": [
          "notes"
        ],
        "properties": {
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/PhonemeNote"
            }
          }
        }
      },
      "PhonemeNote": {
        "type": "object",
        "required": [
          "phonemes"
        ],
        "properties": {
          "phonemes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/Phoneme"
            }
          }
        },
        "description": "The phoneme sequence corresponding to exactly one input note. It contains all phonemes within the selected syllable range, preserving token order and onset flags. phonemes is an empty array when the selected range is empty."
      },
      "Phoneme": {
        "type": "object",
        "required": [
          "token",
          "onset"
        ],
        "properties": {
          "token": {
            "type": "string"
          },
          "onset": {
            "type": "boolean"
          }
        }
      },
      "DurationResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/DurationOutput"
          }
        }
      },
      "DurationOutput": {
        "type": "object",
        "required": [
          "notes"
        ],
        "properties": {
          "notes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/DurationOutputNote"
            }
          }
        }
      },
      "DurationOutputNote": {
        "type": "object",
        "required": [
          "phonemes"
        ],
        "properties": {
          "phonemes": {
            "type": "array",
            "items": {
              "$ref": "#/components/schemas/DurationOutputPhoneme"
            }
          }
        }
      },
      "DurationOutputPhoneme": {
        "type": "object",
        "required": [
          "start"
        ],
        "properties": {
          "start": {
            "type": "number"
          }
        }
      },
      "ParameterResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/ParameterOutput"
          }
        }
      },
      "ParameterOutput": {
        "type": "object",
        "required": [
          "parameters"
        ],
        "properties": {
          "parameters": {
            "type": "object",
            "additionalProperties": {
              "$ref": "#/components/schemas/ParameterOutputParameter"
            }
          }
        }
      },
      "ParameterOutputParameter": {
        "type": "object",
        "required": [
          "values",
          "sampleRate"
        ],
        "properties": {
          "values": {
            "type": "array",
            "items": {
              "type": "number"
            }
          },
          "sampleRate": {
            "type": "number",
            "exclusiveMinimum": 0
          }
        }
      },
      "AudioResponse": {
        "type": "object",
        "required": [
          "output"
        ],
        "properties": {
          "output": {
            "$ref": "#/components/schemas/AudioOutput"
          }
        }
      },
      "AudioOutput": {
        "type": "object",
        "required": [
          "audioUrl"
        ],
        "properties": {
          "audioUrl": {
            "type": "string",
            "description": "Encoded audio URL."
          }
        }
      },
      "AST": {
        "description": "An AST expression is either a numeric literal or an operator expression. Relational and boolean operators interpret non-zero values as true and zero as false; boolean results are represented as 1 for true and 0 for false.",
        "oneOf": [
          {
            "type": "number",
            "description": "Numeric literal."
          },
          {
            "type": "object",
            "description": "Operator expression.",
            "required": [
              "operator",
              "operands"
            ],
            "properties": {
              "operator": {
                "type": "string"
              },
              "operands": {
                "type": "array",
                "items": {
                  "$ref": "#/components/schemas/AST"
                }
              }
            },
            "additionalProperties": false,
            "oneOf": [
              {
                "description": "Zero-operand constants.",
                "properties": {
                  "operator": {
                    "enum": [
                      "x",
                      "e",
                      "pi"
                    ]
                  },
                  "operands": {
                    "minItems": 0,
                    "maxItems": 0
                  }
                }
              },
              {
                "description": "Variadic operators with at least one operand.",
                "properties": {
                  "operator": {
                    "enum": [
                      "+",
                      "*",
                      "hypot",
                      "max",
                      "min"
                    ]
                  },
                  "operands": {
                    "minItems": 1
                  }
                }
              },
              {
                "description": "Binary operators.",
                "properties": {
                  "operator": {
                    "enum": [
                      "-",
                      "/",
                      "%",
                      "pow",
                      ">",
                      "<",
                      ">=",
                      "<=",
                      "==",
                      "!=",
                      "||",
                      "&&"
                    ]
                  },
                  "operands": {
                    "minItems": 2,
                    "maxItems": 2
                  }
                }
              },
              {
                "description": "Unary operators.",
                "properties": {
                  "operator": {
                    "enum": [
                      "abs",
                      "acos",
                      "acosh",
                      "asin",
                      "asinh",
                      "atan",
                      "atanh",
                      "ceil",
                      "cos",
                      "cosh",
                      "exp",
                      "floor",
                      "ln",
                      "log",
                      "round",
                      "sign",
                      "sin",
                      "sinh",
                      "sqrt",
                      "tan",
                      "tanh",
                      "trunc",
                      "!"
                    ]
                  },
                  "operands": {
                    "minItems": 1,
                    "maxItems": 1
                  }
                }
              },
              {
                "description": "Ternary conditional operator: condition, true result, false result.",
                "properties": {
                  "operator": {
                    "enum": [
                      "if"
                    ]
                  },
                  "operands": {
                    "minItems": 3,
                    "maxItems": 3
                  }
                }
              }
            ]
          }
        ]
      },
      "SyllableSliceEndpoint": {
        "type": [
          "integer",
          "null"
        ],
        "default": null,
        "description": "A slice endpoint in the syllable group sequence, expressed as an integer or null; omission is equivalent to null. For M groups, boundary positions range from 0 to M. Nonnegative integers count from the beginning; negative integers first add M, so -1 denotes the position of the last group. Converted positions below 0 are clamped to 0, and positions above M are clamped to M. The start and end fields define their own defaults for null. The start is inclusive and the end exclusive; a start greater than or equal to the end selects an empty range. Reverse selection and a step are not supported."
      },
      "NoteKind": {
        "type": "string",
        "enum": [
          "normal",
          "slur"
        ],
        "description": "normal starts a new pronunciation at this note; slur continues the previous note's pronunciation without introducing new phonemes. A slur note must have an empty phonemes array and gap=0, and cannot be the first note. The slur relationship is expressed by kind, not by the pronunciation string."
      }
    }
  }
}