Okay, so I spent a decent chunk of last weekend training a custom voice in Resemble AI's platform. The results are actually pretty solid for my project's narration needs.
But here's my immediate roadblock: I need to use this voice in a different, locally-hosted TTS pipeline for an offline prototype. I've been clicking around the Resemble web interface for a good 20 minutes and I can't find any "export," "download model," or API endpoint that spits out the actual model file (like a .pth or .onnx). The documentation seems focused on using their API for synthesis, not on model portability.
This feels like a potential lock-in red flag. Has anyone actually managed to extract their trained voice model from Resemble? I'm looking for:
* A hidden setting or a specific API call I missed.
* Confirmation that it's truly not possible, so I know to cut my losses.
* Workarounds—like generating a massive phoneme/audio bank and using that to train another system (painful, but maybe a last resort).
For context, I've used other services where export was either straightforward or at least possible through a convoluted process. I'm hoping Resemble has a path I just haven't found yet.