Open Lidong1211 opened 9 months ago
Hello Lidong1211, and yes, the face render part has not been accelerated. From what I recall, the face render is the part where sadtalker performs data preprocessing, involving a considerable amount of hard-coded image operations. If possible, you might try caching the processing results of identical pictures to reduce processing time.
If it is only targeted to the same video speech synthesis lip, is there more optimization space, please do you have any good suggestions? I'm thinking of lip-syncing for the same video with a different audio.
@kenwaytis
你好,请问faster-SadTalker-API是否对FaceRender部分有速度优化,我看介绍和code仅对FaceEnhancer部分优化了