Nikita Kononov

The problem is that model has 96x96 resolution. So it downscales face square and than upscales to fit your source video. There's no solution. You can only train hi-res model...

> I'm not sure what you mean. The examples you give are exactly how it works at the moment. Is that not enough? See docs/dictionary.md: ' primary stress , secondary...

You can change the network topology to work with 256x256 or 512x512 resolution and train it from scratch

> > No I mean someone else (not me) put together a repo that allows you to fine tune the model. > > Hey, can you send the link of...

@RahulBhalley @hairuo55 @Shivamkumar285 I think I've found it, but not sure mrq/ai-voice-cloning it's not on github, google - first link in search results

> They're just some normal warnings Thank you for your answer. So it doesn't affect inference quality?

> won't have any effect if your model infers correctly Hello! Model converted into onnx with your scripts has very poor performance in NVIDIA Triton Inference Server Inference time is...

> I specified the CPU inference Thanks, I'll give it a try But RunONNX doesn't affect converted model saving, as I can see in the code? I use the converted...

> ok Have done 50 test inferences for each model with same input text pytorch avg ~2.5s onnx avg ~ 2.7s triton onnx avg ~ 4.1s for some reason onnxruntime...

I think, that dropout layer may help. And deeper network topology (more hidden layers in video / audio encoder) Would be glad to see any advice from more experienced enthusiasts)...