Georgi Gerganov

Hey - very cool that you gave it a try and thanks for the feedback! It is relatively easy to make it run asynchronously without blocking the page. I just...

@jaybinks This can be added very easily as a built-in option. A naive algorithm would be for each transcribed segment to measure the signal energy during the time interval for...

> This performance is impressing! Yes, there is a huge performance boost due to using the [built-in BLAS implementation](https://developer.apple.com/documentation/accelerate/blas) on these devices. I will soon add OpenBLAS support for x86...

@trholding Thanks for the results. You can generate a table with performance results by simply running the [extra/bench_all.sh](https://github.com/ggerganov/whisper.cpp/blob/master/extra/bench-all.sh) script. Regarding the threads - yes, it seems that going beyond 8...

Yup - you are missing the `AVX2` flag. See if some of the comments in https://github.com/ggerganov/whisper.cpp/issues/5 can help you resolve this.

I'm not familiar with the WebGPU API. If you demonstrate a basic matrix multiplication example using WebGPU, and it does not look too complicated, I might give it a try.

Can you provide some more details how the "dynamic quantization" works in PyTorch? If it is just converting the weights to 8-bit floating point numbers, then the memory reduction factor...

Yes - there are some tensors from the model that are currently FP32 instead of FP16, because it was easier to first implement the operations in FP32 mode. See this...

Hi, it's possible to port the stream example to WASM, but I doubt the performance is good enough for realtime transcription. By the way, thanks for showing me Vosk -...

Yes, it is possible. By the way, I saw your other comment about Firefox being x5 faster compared to Chrome. I just tried it and I observe similar performance on...