🎙️ Enstruct

Transcribe. Structure. Free.

---

🚀 Overview

Enstruct is a professional, production-ready, open-source audio toolkit that wraps OpenAI's Whisper model (powered by the high-performance faster-whisper engine). Designed to run fully locally, via Docker with GPU acceleration, or on cloud instances, Enstruct offers a free, private alternative to expensive paid services like Otter.ai, Transkriptor, and AssemblyAI.

Key Features

• ⚡ Ultra-Fast Transcription: Powered by faster-whisper utilizing CTranslate2 for up to 4x speedups over standard Whisper.
• 🚀 large-v3-turbo Support: Modern default model size providing high accuracy with significantly lower latency.
• 🌐 Any-to-English Translation: Translate speech from dozens of languages directly to fluent English transcripts.
• 📝 Multi-Format Export: Generates professional SRT and VTT subtitle files, alongside raw TXT transcripts.
• 📂 Batch & Watch Folder Auto-Processing: Transcribe whole folders or continuously monitor inbox directories for automated output.
• 🔌 OpenAI-Compatible REST API: Drop-in API replacement (/v1/audio/transcriptions) for seamless integrations with existing Whisper applications.
• 🐳 Docker & GPU Support: Full Docker Container Toolkit support with NVIDIA GPU passthrough (docker-compose.yml).
• 🎬 YouTube Integration: Transcribe and translate YouTube video content directly via URL input.
• ☁️ Google Drive Storage & History: Persistent session tracking and seamless drive mounting inside Google Colab.
• 🎙️ Flexible Interfaces: Supports CLI, REST API, and interactive Gradio Web UI.

---

📊 Comparison: Enstruct vs Paid Services

Feature	Enstruct (Ours)	Otter.ai	Transkriptor	AssemblyAI
Cost	🆓 100% Free	Paid Subscription	Paid / Minute	Paid / Minute
Hosting	💻 Fully Local / Docker	Cloud-only	Cloud-only	API-only
Privacy	🔒 Zero data shared, fully private	Processed on third-party	Processed on third-party	Processed on third-party
API / Watch Folder	✅ Built-in OpenAI REST API & Watch Folder	Limited	Limited	Paid API-only
Translation	✅ Any-to-English	Limited	Paid	Paid
Open Source	🔓 Yes (AGPL v3 License)	No	No	No

---

🛠️ Installation

pip install enstruct

Note: Enstruct requires ffmpeg to process audio/video container formats.

• Ubuntu/Debian: sudo apt install ffmpeg
• macOS: brew install ffmpeg
• Windows: choco install ffmpeg

---

💻 Usage

Enstruct is designed to work seamlessly as a command-line tool, REST API server, Docker container, or Python library.

1. Command Line Interface (CLI)

# Transcribe an audio file using default large-v3-turbo model into SRT format
enstruct transcribe voice_note.mp3

# Transcribe with target language and VTT format
enstruct transcribe audio.wav --language es --format vtt --output subtitle.vtt

# Watch folder auto-processing (monitors /inbox and saves subtitles to /outbox)
enstruct watch --input-dir /inbox --output-dir /outbox --format srt

# Launch OpenAI-compatible REST API server
enstruct api --host 0.0.0.0 --port 8000

# Batch process an entire directory of mixed audio/video files
enstruct batch /path/to/audios --output /path/to/subtitles --format srt

2. OpenAI-Compatible REST API

Start the REST API server:

enstruct api

Send a transcription request using standard OpenAI Whisper parameters:

curl http://localhost:8000/v1/audio/transcriptions \
  -F file="@speech.mp3" \
  -F model="large-v3-turbo" \
  -F response_format="json"

3. Docker Deployment (Self-Hosted)

Run Enstruct with GPU support using Docker Compose:

docker-compose up -d

This exposes:

• Gradio Web UI: http://localhost:7860
• OpenAI-Compatible REST API: http://localhost:8000

4. Python Library API

from enstruct.core.transcriber import EnstructTranscriber
from enstruct.tools.subtitle import SubtitleGenerator

# Initialize the transcriber (defaults to large-v3-turbo and auto device selection)
transcriber = EnstructTranscriber(model_size="large-v3-turbo", device="auto")

# Transcribe audio file
result = transcriber.transcribe("interview.m4a")

print(f"Detected Language: {result['language']}")
print(f"Transcript: {result['text']}")

# Generate SRT subtitle format
generator = SubtitleGenerator()
generator.generate(result["segments"], "interview.srt", format="srt")

Google Drive Manager

from enstruct.tools.drive import DriveManager

drive_manager = DriveManager()
# Mount Google Drive (Colab) or store locally
drive_manager.save_file("transcript.srt", "Subtitles content...")
# View persistent session history
history = drive_manager.get_history()

---

🌐 Gradio Web Interface

Enstruct includes an interactive browser-based Web UI.

To start the Web UI locally:

python -m interfaces.web.app

Open http://localhost:7860 in your browser. You can:

1. Audio Sources: Upload files, select Google Drive paths, or paste YouTube URLs.
2. Model Selection: Choose models (tiny, base, small, medium, large-v3, large-v3-turbo).
3. Session Logs: View, refresh, or clear transcription history logs.

---

💼 Commercial Use

Enstruct is free under AGPL v3 for open-source and personal use.

If you want to use Enstruct in a closed-source or commercial product, a commercial license is required. See COMMERCIAL_LICENSE.md for pricing and details.

---

🤝 Contributing

We welcome contributions from the open-source community!

1. Fork the Repository.
2. Create your Feature Branch (git checkout -b feature/AmazingFeature).
3. Set up development environment and run tests:

   pip install pytest
   pytest

4. Commit your Changes (git commit -m 'Add some AmazingFeature').
5. Push to the Branch (git push origin feature/AmazingFeature).
6. Open a Pull Request.

---

📄 License

Distributed under the AGPL v3 License. See LICENSE for more information. Commercial licensing and bespoke features are also available.