AI model extension problem, pth to tflite conversion - Coqui XTTSv2

Buongiorno da Roma!

After more than a month, i must give up searching around the web solution and workarounds. I need help from some experts, hoping that you, reading this, know a solution or at least want to try to solve that.

I’have this .pth model:

I have on Unity an integration to make inferences on tflite models, and I MUST use tflites cause the less resources i use for models, the better is. (I need high performance).

Searching and studying i learned that for have a tflite model, starting from a pth, the ONLY way is to convert in this way:

PTH >>> ONNX >>> TF >>> TFLITE

Please, help me with this cause my mind is just blowing up, i don’t know if is really hard or i just need some sort of “AI for dummies” manual. :rofl: :smiling_face_with_tear:

I founded github projects too that should work for this case, but i cannot make it work either from pth to onnx. In case can be usefull for you too or you don’t know a solution but want to try, i drop here all links that i found and actually opened in my browser cause can be a good way to follow, meaby:

https://docs.unity3d.com/Packages/com.unity.sentis@1.4/manual/export-convert-onnx.html
https://pytorch.org/tutorials/advanced/super_resolution_with_onnxruntime.html

Thanks in advance for any contribution, information, test, solution or workaround that you can provide. to obtain this model in .tflite extension.

I have got same problem. Please share your status. What is the current state of conversion? Have you succeeded?

unfortunatly no. Waiting for updates, suggestions, contribution or similar. In the meantime i moved forward on using AI Api TTS from different platforms.