Insanely-fast-whisper support

[lojik-ng]

Please add support for insanely-fast-whisper or faster-whisper

[johnyu0]

Could you please support SenseVoiceSmall and paraformer-zh-streaming ? Wisper’s accuracy in recognizing Chinese is too poor.

[ayancey]

You may consider this project: https://github.com/ahmetoner/whisper-asr-webservice

We do not have OpenAI compatible endpoints yet but working on it!

[matatonic]

I've never used it, but just FYI, this exists: https://github.com/fedirz/faster-whisper-server which supports the openai api.

[matatonic]

FYI: vllm-project/vllm#180
whisper (and encoder-decoder models) to be added to vllm.

I probably wont update this repo much after that, assuming the implementation is good, it would be the preferred way to run open source whisper models with an openai api.