[whisper.cpp](https://github.com/ggml-org/whisper.cpp) is now integrated into auto-editor. This technology allows for transcribing audio, bypassing the need to create subtitles manually.

With a patch set to [ffmpeg](https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21259), word-level transcriptions are possible too.

If you're on MacOS, you can play with these features by downloading [a model](https://huggingface.co/ggerganov/whisper.cpp/tree/main), then run:

```sh
auto-editor whisper example.mp4 ggml-medium.bin
```

You can see the various options via:
```sh
auto-editor whisper --help
```

---

This is still an experimental feature, hence Windows and Linux are not supported yet. I'm also still thinking about how best to integrate these new capabilities in auto-editor proper.