ggerganov
Different platforms return different results
Use the large-v3 model to extract subtitles from the same audio file. Subtitles will be lost under Ubuntu.
./main -m ggml-large-v3.bin -f 1.wav -l zh -osrt 1
In the picture below, Windows is on the left and Ubuntu is on the right.
Also seeing some mild differences between the outcome of large_v3 depending on which CPU whisper is run on (in openvino plugin), AMD and Intel CPUs seem to produces slightly different outcomes even when both are run on WIndows. Differences could be word capitalization, interpretation of things like names and acronyms etc. These are for English.
I just tried the latest version and it is the same on both sides, but the result is wrong and some words are missing.
