- Forum Clout
- 808
The previous thread for context
I finally got some time this week to do a few more RVC models. I made 3, but the Jimmy model sucked, so I'll have to fix it and upload it later. I got sick of going through hundreds of 3-10 second clips of him brooding, so I rushed the dataset and it came out like shit.
Having said that, I wanted to wait until all 3 were done to make this post, but I absolutely had to get this out there ASAP:
David and Bobo 2024
My David impression is meh, but I unfortunately think I nailed Bobo.
These new models can be found here, along with the old O&A ones. I'll probably just use this thread for all future models.
I'm also working on creating TTS autoregressive models to go along with the RVC models, which will allow for some pretty accurate text to speech audio. These TTS models seem to capture the speaking style pretty decently. The data creation and training process is less reliable and more tedious than the RVC stuff, so It's going to take some time before they're ready to be uploaded, but here's a preview:
Again, these are solely for research purposes and not meant to be used with malice or ill intent. Don't do anything illegal.
I finally got some time this week to do a few more RVC models. I made 3, but the Jimmy model sucked, so I'll have to fix it and upload it later. I got sick of going through hundreds of 3-10 second clips of him brooding, so I rushed the dataset and it came out like shit.
Having said that, I wanted to wait until all 3 were done to make this post, but I absolutely had to get this out there ASAP:
David and Bobo 2024
My David impression is meh, but I unfortunately think I nailed Bobo.
These new models can be found here, along with the old O&A ones. I'll probably just use this thread for all future models.
I'm also working on creating TTS autoregressive models to go along with the RVC models, which will allow for some pretty accurate text to speech audio. These TTS models seem to capture the speaking style pretty decently. The data creation and training process is less reliable and more tedious than the RVC stuff, so It's going to take some time before they're ready to be uploaded, but here's a preview:
Again, these are solely for research purposes and not meant to be used with malice or ill intent. Don't do anything illegal.
