README.md

---
title: Video Transcriber
emoji: 🎬
colorFrom: indigo
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
license: mit
tags:
  - transcription
  - speech-to-text
  - whisper
  - video
---

# 🎬 → 📝 Video Transcriber (Speech-to-Text)

Upload a **video** (or audio) file and get a full text transcription, ready to **download** as plain text (`.txt`) or timed subtitles (`.srt`).

## How it works

1. **Upload** a video file — `.mp4`, `.mov`, `.mkv`, `.webm`, `.avi`, `.m4v` (audio files like `.mp3`/`.wav`/`.m4a` work too).
2. Click **Transcribe** — the audio track is extracted with `ffmpeg` (mono, 16 kHz WAV).
3. **OpenAI Whisper** (via 🤗 Transformers) transcribes the audio.
4. Read the transcript and **download**:
   - `*_transcript.txt` — plain text
   - `*_subtitles.srt` — timestamped subtitles

## Options

- **Language** — leave on `auto` for auto-detection, or pin a language for better accuracy.
- **Task** — `transcribe` keeps the original language; `translate` translates the speech into English.

## Model

Defaults to [`openai/whisper-base`](https://huggingface.co/openai/whisper-base) for fast transcription on free CPU hardware. On GPU hardware you can switch to [`openai/whisper-large-v3`](https://huggingface.co/openai/whisper-large-v3) for higher accuracy — just change `MODEL_ID` in `app.py`.

## Tech stack

- [Gradio](https://gradio.app) UI
- [🤗 Transformers](https://huggingface.co/docs/transformers) `automatic-speech-recognition` pipeline
- `ffmpeg` for audio extraction (installed via the Dockerfile)