search

Const-me / Whisper

High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model
Mozilla Public License 2.0
8.2k stars 702 forks source link

generates SRT from video #124

Open lenmao opened 1 year ago

lenmao commented 1 year ago

I'm really perplexed and don't know whether to laugh or cry. Earlier, I used Whisper to convert videos and generate Chinese SRT subtitles from Englisth input, and English SRT subtitles from English input.

However, for the past two days, no matter what I specify, it only generates English SRT subtitles. The audio in the videos is in English. Has anyone else encountered a similar issue?

lenmao commented 1 year ago

I can't quite figure it out now. I have an MP4 video that is over 9 hours long. I'm certain that I used either Buzz or Whisper desktop to convert the English in this video into two separate SRT files: one in Chinese and one in English. However, no matter how I try now, I can't replicate the process, and it seems that Whisper doesn't support this feature. I'm wondering how I managed to do it back then.

emcodem commented 1 year ago

Yes, whisper has not been trained to translate from english to chinese but it can by accident output chinese subtitles for english audio in case it has accidently been trained "with wrong data". IF whisper translates english to chinese, it is accident or better wrong trainging; a consequence of the weak supervised training according to the original documentation