Fictionarry / ER-NeRF

[ICCV'23] Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis
https://fictionarry.github.io/ER-NeRF/
MIT License
906 stars 125 forks source link

生成的视频没有声音 #56

Closed yq-xfl closed 9 months ago

yq-xfl commented 9 months ago

我训练完后,用音频进行测试,口型大致能对上,但是results里的mp4没有声音,可能是哪里的问题呢?

还有如何我用更大的分辨率,比如1024*1024的视频,是否可以进行训练呢?

Fictionarry commented 9 months ago

没有声音是正常情况,见https://github.com/Fictionarry/ER-NeRF/issues/9 目前的超参数不支持1024分辨率的训练,需要自己调整参数和结构