https://github.com/mbotsu/mlx_speech2text

Audio transcription using mlx whisper and vad silence processing
https://github.com/mbotsu/mlx_speech2text

mlx silero-vad whisper

Last synced: about 1 month ago
JSON representation

Audio transcription using mlx whisper and vad silence processing

• Host: GitHub
• URL: https://github.com/mbotsu/mlx_speech2text
• Owner: mbotsu
• License: mit
• Created: 2024-07-08T12:30:28.000Z (4 months ago)
• Default Branch: main
• Last Pushed: 2024-08-14T13:59:19.000Z (3 months ago)
• Last Synced: 2024-09-27T06:22:13.183Z (about 2 months ago)
• Topics: mlx, silero-vad, whisper
• Language: Python
• Homepage:
• Size: 13.7 KB
• Stars: 7
• Watchers: 1
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
## Abstract

Transcription for Apple Silicon.

Segmentation is performed to divide the sound source into small chunks, a sound source is created by removing silent parts for each chunk, and text is extracted.

## Install

```

$ git clone https://github.com/mbotsu/mlx_speech2text.git

$ pip install -r requirements.txt

```

## Run

```

// convert to wav 16K

$ ffmpeg -i input.mp4 -ar 16000 out.wav

// run

$ python speech2text.py -i out.wav -o track -v

```

## References

- [ml-explore/mlx-examples/whisper](https://github.com/ml-explore/mlx-examples/tree/main/whisper)

- [snakers4/silero-vad](https://github.com/snakers4/silero-vad)

- [Softcatala/whisper-ctranslate2](https://github.com/Softcatala/whisper-ctranslate2)

- [Segmenting a long audio file #295](https://github.com/snakers4/silero-vad/discussions/295)