Fast audio/video transcription using Whisper, optimized for NVIDIA GPU and Apple Silicon.
curl -sSL https://raw.githubusercontent.com/aaditagrawal/transcript-cli/main/install.sh | bashRestart terminal, then use.
uv pip install git+https://github.com/aaditagrawal/transcript-cli.git
uv pip install mlx-whisperJust run with a file - you'll be prompted for options:
transcript video.mp4You'll see:
Select output format [text/timestamps/srt/vtt/json]: srt
Select model size [tiny/base/small/medium/large-v3/turbo]: base
Specify options directly:
transcript video.mp4 -f srt -m base # SRT subtitles with base model
transcript video.mp4 -f json -m large-v3 # JSON with large model
transcript ./folder -r -f vtt # Batch process folder| Format | Flag | Description |
|---|---|---|
text |
-f text |
Plain text transcript |
timestamps |
-f timestamps |
Text with [HH:MM:SS] markers |
srt |
-f srt |
SubRip subtitles |
vtt |
-f vtt |
WebVTT subtitles |
json |
-f json |
Full metadata + word timestamps |
| Model | Flag | Speed | Quality |
|---|---|---|---|
tiny |
-m tiny |
Fastest | Basic |
base |
-m base |
Fast | Good (default) |
small |
-m small |
Medium | Better |
medium |
-m medium |
Slow | Great |
large-v3 |
-m large-v3 |
Slowest | Best |
turbo |
-m turbo |
Fast | Great |
# Quick subtitle generation
transcript movie.mp4 -f srt -m base
# High-quality transcription
transcript interview.mp3 -f json -m large-v3
# Process all videos in folder
transcript ~/Videos -r -f srt -m turboMIT