Updates for version 4.x

[subof]

Hi,
We are certainly happy that the new version is getting new features, but could you please note that version 4.15 contains more familiar tools and could be developed until the new version is ready to replace it?
In batch mode, version 4.15 contains more options, and some of them could be improved. The auto-balance option in batch mode does not contain the same settings as the right-click option. Could you please add this setting to batch mode? Version 5.x seems to have fewer settings.
Additionally, it would be convenient to have new TTS tools in version 4.x.
We are really looking forward to the new 5.x version, and we will be happy if the 4.x version continues to receive updates for a while.
Thank you.

[Ironship]

@subof as far as I know SE4 is in debug mode. Bugs will be fixed but new feature requests should be to SE5. Changes between SE4 and SE5 are large so doing new features for both adds lots of overhead which should be used for polishing SE5.

[niksedk]

@subof: What are you missing in SE 5?

[subof]

@subof: What are you missing in SE 5?

It sometimes freezes up and I have to remove it from the processes, and as I said, there are fewer settings, at least in batch mode.

One of the wishes, as I wrote earlier, is to set up a configuration for specifying the generation folders. Currently, this is unmanageable, and all tests end up in the system directories.
Qwen generates well, but it has problems with accents, and it's unclear how to manage them. The standard approaches often don't work: +/`/cAps.
Chatterbox is controlled by the standard accent mark (´).

My goal is to achieve synchronous dubbing, where the model generates speech not based on a single sample, but rather on the current prosody. After all, the model doesn't know whether the character is whispering or shouting at a particular point in the text. A single sample is not sufficient. Your program's achievement lies in using timings, which allows you to read based on time stamps, something that is rarely done. We live in a world where time is the fourth dimension, and ignoring it is unnatural.
How could I dub subtitles for a movie with multiple characters?
Perhaps you could create a module where I can specify the characters in advance, or the program could dynamically sample the entire audio track to match the character's emotions at each moment.
You already have an audio waveform and the ability to visually edit it, which I believe could serve as a starting point.

[niksedk]

I will close this for now - feel free to create a request issue for a feature you are missing from SE4 (or new request).
Speaker recognition might come later when/if an AI model for that appears (and is easy integrated).