whisper-mps

An opinionated CLI to transcribe Audio files (or YouTube videos) w/ Whisper on-device! Powered by MLX, Whisper & Apple M series

TL;DR - After our actual testing, the Whisper supported by MPS achieves speeds comparable to a 4090!

80 mins audio file only need 80s on APPLE M1 MAX 32G! ONLY 80 SECONDS

🆕 Blazingly fast transcriptions via your terminal! ⚡️

We've added a CLI to enable fast transcriptions. Here's how you can use it:

Install whisper-mps with pip:

# please install ffmpeg first: brew install ffmpeg
pip install whisper-mps

Run inference from any path on your computer:

# filetype should be wav/mp3/mp4 etc.
whisper-mps --file-name <filename>

Run inference with a specific model size:

# for example, using the base model size. Other available models: "tiny", "base", "small", "medium", "large".
# Larger models require more loading time.
# filetype should be wav/mp3/mp4 etc.
whisper-mps --file-name <filename> --model-name base

Run inference from a YouTube URL on your computer:

[!NOTE] Deprecated: The YouTube URL inference feature is deprecated and may be removed in future releases.
Please uses yt-dlp to download YouTube videos.

# whisper-mps --youtube-url https://www.youtube.com/watch\?v\=jaM02mb6JFM

[!NOTE] The CLI is highly opinionated and only works on Apple MPS.

CLI Options

The whisper-mps repo provides all-round support for running Whisper in various settings. More command-line support will be provided later.

  --file-name FILE_NAME
                  Path or URL to the audio file to be transcribed.
  --model-name MODEL_NAME
                  Size of the OPENAI Whisper model name, like tiny (default), base, small, etc.
  --youtube-url URL_ADDRESS
                  The YouTube video URL. (Deprecated)
  --output-file-name OUTPUT_FILE_NAME
                  The output file name for the transcribed text JSON.
  --log-level LOG_LEVEL
                  Optional. ["DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL"]