The new version Subtitle Edit with the addition of Whisper Faster to the menu (I guess) just dropped. I have no way of knowing if special hardware is required for Whisper Faster and run on Win10 and the CCp (whatever) parameter.
I'll try it out on something short but that is not the test worth doing. As stated previously, the Whispers have given dropouts when going past a musical passage as example. It will run, hit the [Music] and not return to analyzing voice.
+ Reply to Thread
Results 1 to 6 of 6
Thanks for letting us know about the new version.
Subtitle Edit got a 4.01 upgrade the last couple days but not related much to voice to text and no improvement either on the Whisper handling of
what? I can't follow it -- all of a sudden there's a whole screen of lines of duplicate text. It's easier to do a quick inspection in an SRT reader like Notepad++ so I didn't even try to watch any of it. All that cannot be dependent on GPU to function. If so, then how could the program checks of the model be _slowed down_? Just wondering.
I do wonder what all the improvements cited to various language translation dictionaries are in each release of SE and how they get implemented. Again, just a mystery since no one tests any of these features.
With all these Whisper problems what is going on with VOSK-- anything? New Models? updates?
A partial touche'
I did try once I think with a friend from Poland but we're not real conversant on this topic so I think
that part of my the message asking to participate in a test from a native speaker was ignored.
There is a translate group on Reddit of native speakers FIY. They've given me some assistance.