4
Transcribing lectures using Whisper.md
```sh
whisper --model medium.en --language English --output_format vtt --fp16 False --word_timestamps True --max_words_per_line 30 --initial_prompt "<prompt>" <input-file>
```
 
Offline transcription is roughly real-time (i.e., 1 hour of audio takes about 1 hour to transcribe).
 
Models are downloaded to `~/.cache/whisper`.
 
`whisper-cpp` is actually the one we want as its written in C++ and supports Core ML. Annoyingly the CLI options are different, but it seems to have more of them. Uses the same models as Vibe below. Input audio must be 16kHz 🙁 (`ffmpeg -i <input> -vn -ar 16000 <output>`).
 
```sh
whisper-cpp --model ~/Library/Application\ Support/github.com.thewh1teagle.vibe/ggml-medium.en.bin --language en --output-vtt --max-len 150 --prompt "<prompt>" <input file>
```
 
**Much** faster.
 
**Vibe** seems to be a useful cross-platform GUI implementation. Internally whisper-cpp ported to Rust. Seems to produce very short line lengths. it claims to have a CLI, but I cant figure out how to make it work, and the max sentence length setting doesnt seem to help (ohhhh, its measured in **characters**, not wordsduh 😖). It produces not-quite-correct VTT: no WEBVTT header, and no blank line between entries.
 
VS Code extension issues: