Georgi Gerganov

When I implemented the x2 speed-up option I did a small research on tempo speed-up algorithms and it looks like the general solution is not very trivial to implement because...

@abitofevrything Can you verify that everything is good on Windows? After that I will merge it

The `whisper_full_params` struct has to be mapped precisely. I immediately see that the order of the `n_threads` and `n_max_text_ctx` is wrong: 

@alex-bacart The `--max-len 1` means to output maximum 1 token per text segment. The word " Ponzi" consists of 2 tokens: ` Pon` and `zi` and therefore it is being...

@mightymatth Thanks for this contribution - I think this is very useful! Although it is OK to merge like this, I will likely change it to have a bool flag...

Thanks for the data point! How do I calculate WER scores?

Hi, thanks for the detailed steps - this helps a lot. After debugging with [WHISPER_DEBUG](https://github.com/ggerganov/whisper.cpp/blob/b2083c5d02db9a1e6dbb3d58254fd65ebfff4b5d/whisper.cpp#L91) enabled I can see immediately that in this case, the entropy-based check for repetition didn't...

There might be issues on 32-bit platforms - need to see if there are some 64-bit assumptions in `ggml.c`. Regarding the Makefile - it is very likely that the current...

Yup, logging settings will be exposed through the API at some point and will solve similar issues. Btw, currently whisper prints everything related to internal processing to STDERR, not STDOUT....

This way, the stream example will keep on auto-detecting the language forever. Maybe it would be better after it auto-detects the spoken language to select it explicitly in order to...