Transcribe Tool

Local audio/video transcription using Whisper. Produces TXT, SRT, VTT, or JSON and runs fully offline. Use the CLI for scripting or the web GUI for a point-and-click experience.

Requirements

• Python 3.10+
• ffmpeg on PATH

Install

pip install -r requirements.txt

ffmpeg:

• Windows (winget): winget install --id Gyan.FFmpeg -e
• macOS (brew): brew install ffmpeg
• Linux (apt): sudo apt-get install ffmpeg

Usage

python transcribe.py INPUT [--model MODEL] [--language LANG]
                           [--format {txt,srt,vtt,json}] [--output PATH]

GUI

Launch the web interface:

python app.py

Opens a browser tab at http://localhost:7860.

1. Upload an audio or video file
2. Pick a quality level — Fast (tiny), Balanced (small), or Best (large-v3)
3. Click Transcribe

Advanced options (output format, language) are available under the expandable accordion. The transcript appears in a preview pane with a copy button, and a download link is provided for the output file.

Flags

• --model: Whisper model name. Examples: tiny, base, small, medium, large-v3. Default: small.
• --language: Language code like en, es. Omit to auto-detect.
• --format: Output format: txt, srt, vtt, json. Default: txt.
• --output: Output path. Defaults to input name with chosen extension.

Examples

# Basic transcription to TXT
python transcribe.py sample.mp3

# Force language and output SRT
python transcribe.py sample.mp4 --language en --format srt

# Use a larger model and custom output file
python transcribe.py sample.wav --model medium --output out/transcript.txt

Notes

• Both the CLI and GUI prefer faster-whisper and fall back to openai-whisper if needed.
• CLI output is written next to the input unless --output is provided. GUI output is available via the download link.
• Recommend using the tiny/Fast model for larger audio files or low CPU availability.