Silicon-Transcription AI

My own scripts to transcribe from a terminal some audio files.

Pre-requisites

Check ffmpeg and ffprobe are installed and available.

ffmpeg -version
ffprobe -version

Install the env and get the whisper.cpp model.

uv sync
git clone https://github.com/ggerganov/whisper.cpp
cd whisper.cpp
sh ./models/download-ggml-model.sh large-v3
cmake -B build -DWHISPER_COREML=1
cmake --build build -j --config Release

Usage

• Transcribe a local audio file.

uv run stai transcribe --model large-v3 --file_path /path/to/audio.mp3

• Transcribe a Youtube video.

uv run stai transcribe --model large-v3 --url https://www.youtube.com/watch?v=dQw4w9WgXcQ

• Transcribe a URL.

uv run stai transcribe --model large-v3 --url https://example.com/audio.wav