FR: compatibility with OpenAI's API

[didmar]

If I'm not mistaken, this API is different from OpenAI's.
It prevents usage as a drop-in replacement for other applications or plugins that connect to Whisper (such as https://github.com/nikdanilov/whisper-obsidian-plugin).
This would be great to avoid ad-hoc implementations in each of these applications and plugins.

[ayancey]

I like this idea. It would be useful if you could explain OpenAI's endpoints a bit and what the output format is for the data. thank you

[didmar]

I've ended up forking a project that does specifically that: https://github.com/didmar/whisper-api-server
There is also the official doc: https://platform.openai.com/docs/guides/speech-to-text

[ahmetoner]

Good idea! I was actually planning to implement this change in version 2.0. I'll start to make some breaking changes to the repository after the new year.

[fedirz]

I've recently created faster-whisper-server, which supports /v1/audio/transcriptions, /v1/audio/transcriptions, /v1/models, and /v1/models/{model} routes, as well as streaming transcription/translation (SSE), and live audio transcription (web socket)

[ayancey]

proof of concept: #237. Still needs a lot of love and some abstraction to work across openai-whisper/faster-whisper cores