Open ando818 opened 6 months ago
Would really love an answer as to how to optimize this
AFAIK Voicemod doesn't use any ML or AI models, it's just plain algorithms of audio processing (echo, pitch, distortion etc.) that's why it's so fast
@mlatysh Interesting, that's my suspicion but they advertise AI voices all over the places. Though entirely possibly they mean something else entirely
The paper says "low-resource voice conversion". Using the provided infer script, my CPU usage is at 55% with a Ryzen 7 3800x 8 core
Is there any reason the inference script is high on CPU? Are there any quick optimizations to be made here? Voicemod is achieving real time conversion with 8-10% cpu usage in comparison, so I was expecting this to work similarly or better