Telegram Subtitle Generator Bot

Python bot that receives Telegram video/audio files and returns .srt subtitles using Google Speech-to-Text.

Features

• Accepts video, audio, voice, and media document uploads.
• Validates:
  • max file size: 50 MB
  • max duration: 20 minutes
• Converts media to mono 16kHz WAV with FFmpeg.
• Transcribes in chunked mode to stay memory-friendly on low-resource hosts.
• Generates .srt with timestamps.
• Enforces single active heavy job.
• Aggressive cleanup:
  • source file deleted after extraction
  • temp folders cleaned at job end
  • stale temp folders auto-purged by TTL sweeper (default 120s)

Required Environment Variables

Use .env.example as reference.

• BOT_TOKEN
• WEBHOOK_URL (for Render deployment)
• WEBHOOK_SECRET_TOKEN (recommended)

Note: Google Speech-to-Text still requires valid Google Cloud authentication at runtime (for example via platform-provided default credentials).

Built-in defaults (no env needed):

• webhook path: /webhook
• language: en-US
• max file size: 50MB
• max duration: 20 minutes
• temp TTL cleanup: 120s

Local Run

pip install -r requirements.txt
uvicorn app.main:app --host 0.0.0.0 --port 10000

Render Deploy

1. Create a new Render Web Service from this repo.
2. Use Dockerfile (or render.yaml Blueprint).
3. Add all required env vars.
4. Set bot webhook URL env to your service URL:
   • WEBHOOK_URL=https://<your-render-service>.onrender.com

Webhook endpoint:

• POST /webhook

Health endpoint:

• GET /health