Match-TTS Onnx Benchmarks
Click button to load a model
almost load time 15 sec,infer time 2 sec(my 2070super-gpu)
Quantized version is too slow and exist just for Github Page 100MB limitation so far
Multispeaker(vctk) is littlebit slow than singlespeaker.en001 is littlg bit firster maybe timesteps is 5(bigger hq but slow)