Roger Peng

运行非常慢，不到2分钟的音频，运行超过12个小时进行Face Renderer，最后因无法分配高达1.26GiB的内存空间而产生异常崩溃了。

2

E:\Software\SadTalker>D:\ProgramData\Anaconda3\python E:\Software\SadTalker\inference.py --driven_audio E:\Software\SadTalker\audio\NewsAboutAItoday.wav --source_image E:\Software\SadTalker\img\PrettyGirl5.jpg --result_dir E:\Software\SadTalker\output --still --preprocess full --enhancer gfpgan E:\Software\SadTalker\checkpoints\epoch_20.pth E:\Software\SadTalker\checkpoints\auido2pose_00140-model.pth E:\Software\SadTalker\checkpoints\auido2exp_00300-model.pth E:\Software\SadTalker\checkpoints\facevid2vid_00189-model.pth.tar E:\Software\SadTalker\checkpoints\mapping_00109-model.pth.tar 3DMM Extraction for source image landmark Det:: 100%|███████████████████████████████████████████████████████████████████████████| 1/1 [00:15