Just to be clear, I am not trying to clone or copy my voice or anyone else's. When feeding Suno Bark text for generation, it will attempt to generate a voice that matches the text. It does this very well. However, the utility is limited. There are preset voices in the form of npz files you can use, but they are fairly dull and unexpressive compared to the randomly generated voices. Further, there is only one preset with a female English voice.
There must be a way to capture the seed or state from a generated voice, so as to use that seed or state in future generations. The goal is to create new presets based on previous generations, or save the state so it can be reused in longform generation. I'm not sure if this involves creating a new npz file or what. Has anyone successfully accomplished what I'm trying to do?
