Are VoiceDesign /Cloning viable?
May be as simple as combining existing voices? Or Lora?
yes, zero-shot voice cloning is supprted!
voice design isn't supported yet.
Can cloned voice be stored for later use? Is mixing/voice design planned, and if it is - in what timeframe?
yes you can reuse a cloned voice and use it for reference later. lmk if you're having trouble doing this and i can provide more detailed instructions.
voice design should be available in about 1-2 months.
Please do. I mean not reuse whole process of distilling from source, more like saving distilled tensors and reusing them after.
You can already run zero-shot voice cloning. You don't need to save distilled tensors for that. you just need to store the speaker embedding and the reference audio for this. Please let me know if I'm misunderstanding something
If you can be so kind, please, post the whole procedure. It's like I don't understand something. If it's already posted somethere, as a doc instruction, the link would be great ( model card short doc is not enough to comprehend it, for me. I see only wav -> cloned voice instruction, but in my understanding, it's seed-related and has big overhead beyound sample wav -> save clone result -> generate new, via saved result (each time loading result) pipeline.
I'm not native speaker, so please forgive me if I'm not explaining meaningfully.
m.generate("[joyful] We won the grant I can (((hardly))) believe it!",
voice="Kiki", preset="expressive")
Like with "voice" param here.
By the way, does presets work on cloned voices?