Const-me
Request: word-level timestamps
The word-level timestamp is a very interesting and useful feature introduced in the recent version of OpenAI's Whisper. It not only makes the subtitle display more synchronized with the voice but also displays a flashing horizontal line at the bottom of the subtitle of the currently spoken word during playback, which is very helpful for language learners to concentrate.(https://github.com/openai/whisper/commit/500d0fe9668fae5fe2af2b6a3c4950f8a29aa145)
This was actually already implemented by gerganov's whisper.cpp. You can specify the character length of each line ( -ml 40 , to get 40 characters) and when you get down to 1 it will switch over the word level. I thinking this repo would just have to expose the feature. Would be extremely useful!
这实际上已经被格尔加诺夫的耳语.cpp实现了。您可以指定每行的字符长度（-ml 40 ，以获得 40 个字符），当您降到 1 时，它将切换单词级别。我认为这个回购只需要公开该功能。将非常有用！
桌面版的可以指定吗
这实际上已经被格尔加诺夫的耳语.cpp实现了。您可以指定每行的字符长度（-ml 40 ，以获得 40 个字符），当您降到 1 时，它将切换单词级别。我认为这个回购只需要公开该功能。将非常有用！
Can it be specified for the desktop version?