Curious of OuteTTS Usage

[lenankamp]

Been playing with the OuteTTS implementation and curious about the custom input. Just been cramming character descriptions in the field and getting a hashed seed that's at least distinct for each speaker. Can't say I've heard a random voice yet superior to any of the normalized presets, not to say there weren't some interesting results.

Mostly curious if others have any tips for best results. This approach is giving me at least distinct voices for speakers, but any tips for refinement would be appreciated.

[LostRuins]

In the next version 1.87, voice cloning will become available. This will give you much better control over the voices.
You can sneak a peek here https://github.com/LostRuins/koboldcpp/blob/concedo_experimental/examples/outetts/