Georgi Gerganov

Hmm, not sure what is the problem. Try to compile a basic C++ program and see if it works: ```c++ // test.cpp #include int main() { return 0; } ```...

Looks like a bug - can you provide an audio sample?

This looks very well done - thank you for the contribution! Regarding the need for the `WHISPER_NO_AVX2=ON` CMake option - maybe if you use `make libwhisper.a` instead of CMake it...

There is a check to stop processing if less than 1s of audio remains: https://github.com/ggerganov/whisper.cpp/blob/a750868428868abd437e228ae5cab763ef3dc387/whisper.cpp#L5271-L5277 I've figured it helps in most situations, but obviously can affect negatively in cases like...

Can't reproduce on MacOS

Interesting question. I am also curious if one can devise a strategy that works for words or multi-token strings, ideally without having to backtrack. I guess some form of rejection...

I'll look in more details into the threading work after I finish with the quantization improvement efforts and some other pending stuff. But at the moment, I can immediately say...

I don't have a `.clang-format` - it would be great to create one that matches the existing style if possible

I think I've seen a tool that can analyse a codebase and try to generate `.clang-format` that is very close to the existing style. I might be wrong though

This fell out of priority - feel free to give it a try