Please forgive me if I don’t respond quickly; I’ve been having internet connection issues since yesterday and am waiting for it to be solved.
For those of you who have asked me to send you the beta, no problem, but please wait until I’ve solved the problems that have been reported. I’m working on it.
I should clarify that the problems are focused on the installation, not on the operation of the application itself. It’s not a “simple” single installer; rather, it installs several dependencies, and some of them behave in one way or another depending on the computer they are installed on. I’m trying to solve this so that I can have a more robust installer file that is better prepared for failures.
After struggling a bit to solve the installation problems, I now have a new version ready…
Changes:
Now works with PyTorch 2.7 + CUDA 12.8, which is more modern and therefore compatible with Nvidia 30xx, 40xx, and 50xx GPUs (and probably earlier ones).
Natten and All-in-One versions updated to work with PyTorch 2.7+CUDA 12.8
Local “wheels” to reduce potential installation problems (no longer requires downloading certain dependencies, they are inside the “\wheels” folder, ensuring that they are the appropriate versions in each case for the application and the “venv” environment created)
New .bat files: there is now one .bat file for installation/verification and another .bat file just for starting the application
Improvement: Now, if the audio/video file is not in WAV format (e.g., it’s in MP3 format), that audio is always converted to WAV and used for all the different processes, avoiding potential offset issues.
As I’m still having problems with my fiber optic connection, I can’t upload the new version right now, it would take too long. I hope that tomorrow the fiber optic connection will finally be fixed.
In the meantime, I’m taking the opportunity to fix other things:
Review pending translations into English.
Preventing unnecessary wav2vec2 models from running in “AI Analysis” (this is a recent problem, I hadn’t discovered it until now, and it’s due to the latest mood I added, “Fap Mixer”).
Fixing a playback issue in “Point Editor.”
Fixing an audio processing issue due to the new version of allin1fix.
The new version is now uploaded, and I have sent the link to the three people who helped me and reported problems/bugs in the previous version. Thank you all!
As soon as they confirm that it works well for them, I will send it out to the rest of you
New mood created, called “SixthSense” for a reason, lol
It uses separation with another AI model that allows 6 stem extraction: Drums, Bass, Other, Voice, Guitar, and Piano.
It also uses drum separation for: Kick, Snare, Toms, and Cymbals.
It allows great variety in the up-down range as well as adaptation of the amplitude between them.
It also allows cool things like activating vibrato in guitar solos. Oh yeah!
I was able to test the funscript generated by the “SixthSense” mood with The Hadny and… WOW! I liked it! LOL
In a couple of hours, I’ll upload the video and the funscript so you can try it out. I want to add some “Edge & Cum” blocks at the end to make it more epic, lmao
I liked it! 
