prescription
For users with limited video memory, the following methods can be used to solve the video memory problem:
- Choose an efficient configuration:Using the stage2_efficient.yaml configuration file
- Reduced resolution:Inputs and outputs maintain 256 x 256 resolution
- Reduce the number of inference steps:Adjust the inference_steps to about 20
- Batch processing:Avoid running multiple instances at the same time
With the above optimizations, a 6.8GB video memory card can meet the basic running requirements.
This answer comes from the articleLatentSync: an open source tool for generating lip-synchronized video directly from audioThe
May not be reproduced without permission:AI productivity tools " How to fix LatentSync running low on video memory?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil