Alex Jones (Infowars conspiracy nutjob)
Louise Belcher (Bob's Burgers)
Solid Snake (Metal Gear Solid 3 so Naked Snake technically)
(it was trained on this 1:14 sample https://vocaroo.com/1bKjtrkZCWSO
dataset used -> https://files.catbox.moe/csbnwp.rar )
Harry Mason (Silent Hill)
James Sunderland (Silent Hill 2)
Mary Sunderland (Silent Hill 2)
Melina (Elden Ring)
Mitsuru Kirijo (Persona 3)
(You might have to rename the cond_latent to "cond_latents_f3c7b7fb.pth")
and a dataset for Sersha from /gif/ (the poster don't wanna share his model for some reason)
A discussion with some shared models https://git.ecker.tech/mrq/ai-voice-cloning/issues/60
How to use:
Use with Tortoise mrq fork https://git.ecker.tech/mrq/ai-voice-cloning/wiki/Installation
Place the finetuned model .pth file (1.6GB) in models/finetunes.
Select it in Gradio UI: Settings tab -> "Autoregressive Model".
If a cond_latents_[hash].pth file (<2MB) was provided, put it in your voices/[voice]/ folder, you don't need audio samples.
If not, use regular audio samples in your voices/[voice]/ folder.