search

KdaiP / StableTTS

Next-generation TTS model using flow-matching and DiT, inspired by Stable Diffusion 3
MIT License
348 stars 39 forks source link

How is the performance of inference speed? #5

Open GavinZhao19 opened 6 months ago

GavinZhao19 commented 6 months ago

Are there any specific data, in terms of performance on CPU and GPU?

KdaiP commented 6 months ago

Thank you for your interest in the project

I'm able to achieve a speed of approximately 200 tokens per second on my own computer's CPU. I plan to release the pretrained model within the next one to two weeks, at which point you'll be able to conduct your own tests.

GavinZhao19 commented 6 months ago

Thank you for your interest in the project

I'm able to achieve a speed of approximately 200 tokens per second on my own computer's CPU. I plan to release the pretrained model within the next one to two weeks, at which point you'll be able to conduct your own tests.

Great work! Achieving 200 tokens per second on the CPU is impressive. I'm looking forward to the pretrained model you've trained.