Open yuvalBGU1 opened 2 months ago
It's not the voice cloning that adds to the loading time, it's loading the base XTTS model. That occurs everytime you use anything that uses coqui XTTS, independent from RealtimeTTS, it's just part of the tts library. Once the model is loaded (CoquiEngine created) you can fast switch voices with the set_voice method. But nothing I can do to speed up the initial loading time of base XTTS.
Hi, i was wondering if there is an option to save a model with a specific voice to shorten the loading time, for example, i now load the v2.0.2 xtts model with a voice path: engine = CoquiEngine(voice="vocals/id1/tr01.wav") every voice has its own .json file, and i need some option to shorten the loading time, perhaps with the ability to load a pre-configured model. Thanks for the help!