Georgi Gerganov

@trappedinspacetime You can try adjusting the VAD-related parameters: ``` -vth N, --vad-thold N [0.60 ] voice activity detection threshold -fth N, --freq-thold N [100.00 ] high-pass frequency cutoff ``` Probably...

@bobqianic We should better have the batched decoding implemented before additional tests. Without it `whisper.cpp` will always be significantly slower

How much is the `encode time, ms per run` without `WHISPER_COREML`? Check only the first run, as in this screenshot: 

Try using the `stream` example with the `--lang auto` argument. Not sure if it will work, but might be worth giving at a try. In general, switching languages is not...

@janekb04 This is very nice work! I've always had doubts that the existing spin-lock approach is not optimal, but my attempts of adding mutex and condition variables were giving worse...

> I just saw that `ggml.c` is copy-pasted to `llama.cpp`. I'll see if it improves performance there. The `ggml.c` in `llama.cpp` has some new extra stuff added and I haven't...

The fallback implementation currently is suboptimal and I think this is causing the slow performance. Try using `--no-fallback` for now, and in the future we will try to improve the...

Can you confirm that adding `-nf` resolves the issue?

The fall back generally improves transcription quality. But currently, the evaluation in `whisper.cpp` is suboptimal. I plan to disable this option by default until more efficient implementation is available

Does `whisper-cpp` work for executable name?