@DaisyAzuras
This specific one was generated by a paid model (Google Gemini), but I think the latest minimax release could do this on a local setup if you have enough vram. I have another video of Scarlet walking down a bridge that was done with minimax if you want comparison.

@Scarlet Ribbon
what is her personality like?

how do yo get results like this? is this generated locally or on a server?

@Bad Dragon
TBH, given her personality… I can’t elaborate on that because of the rating of this image.

That’s quite a phallic-shaped object she’s engineering there. Is she sure this would be safe to use?

@Scarlet Ribbon
I really want something better than Voicemod that doesn’t sound glitchy and autotuned. Gemini tells me that Elevenlab’s voice creator is the best option for that (though it is subscription based). Maybe another time I’ll get around to researching it.

Basic goal: Find a way to generate some clean audio clips of a voice that doesn’t exist, make 15 minutes of lines from it, and then train on RVC.

@Meta_Void
I have actually had these same thoughts. I have an idea of what Scarlet SHOULD sound like in my head, but I’ve never been able to ‘hear’ her, if that makes any sense. I’d love to get that sorted out, cuz I have around two dozen heavily developed OCs I’d love to bring entirely to life.

@Scarlet Ribbon
This just reminded me of one of the biggest hurdles I have faced which is: How do you give an OC a custom voice? And at this point, how do you make a custom voice work with a model like this? Perhaps if you made a custom AI voice with something like voicemod, and then recorded a bunch of lines with it, maybe you could train that data on something like RVC in order to use it forever?

@Scarlet Ribbon
Aha. Well I really like the OC, and the drip.

@Meta_Void
The prompt didn’t include any dialogue and it generated with nothing but soft music, so I stripped the audio for filesize reasons. Actually, I didn’t even prompt the blueprint: I was doing design discussion with Gemini and it remembered that she’s an engineer.

Lol at the hair swapping around like in the show. Where audio? If you train a voice on an audio to audio speech model you can use openvino audacity to seperate the stock voice from the background noise and put in whatever voice you want while keeping the lipsync.