MuseTalk

MuseTalk copied to clipboard

Making MuseTalk 40% faster

8

I've been pretty impressed with MuseTalk albeit some of its shortcomings and have been playing around with the model. Ended up doing a ton of optimizations that made it run...

mvoodarla

I generated this video where the Lip sync is good but low in resolution. Adjusting the bbox parameter doesn't help. Can someone please help resolve this? https://github.com/user-attachments/assets/8b001342-f693-4a30-bef3-0c7d77f1752e

gokula-krishna-dev

为什么合成到最后报一个错，FileNotFoundError: Input video file not found: ./temp.mp4

3

pad talking image to original video Traceback (most recent call last): File "D:\ProgramData\miniconda3\Lib\site-packages\gradio\queueing.py", line 536, in process_events response = await route_utils.call_process_api( ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "D:\ProgramData\miniconda3\Lib\site-packages\gradio\route_utils.py", line 276, in call_process_api output =...

dahaixingchen

你好，人脸处理完之后，很多细节丢失了，有没有参数或者方法保留细节

11

处理前人脸：  处理后人脸： 

raymondren1982

试了一圈，wav2lip、video-retalking、geneface++、musetalk等等，感觉没有一个能直接落地的

15

那些商用的，号称使用5分钟视频就可以定制，有大佬知道方案么？

myhostone1990

预处理阶段视频拆帧

2

请问保存在results里面的帧，训练时还会用到嘛，我用多个视频进行训练的话，预处理阶段会保存所有视频的帧，占用大量磁盘空间，如果待提取人脸结束，我在哪里删除帧呢，非常感谢您的回答！

miumiuc

Support for other whisper models specifically Whisper-medium

1

The problem with using whisper models other than tiny is that the pre-trained models/weights do not work properly with other versions. The encoder and checkpoint functions cause problems due to...

devs-hooked

内存释放问题

4

/root/anaconda3/envs/muse/bin/python 7624MiB 启动运行过后 gpu一直在7g左右，请问哪里可以优化释放内存. 里面unet 和 vae 用的么？

mmlingyu

动物的口型驱动

2

动物的口型识别这块有问题 有计划支持么，或者有建议的方法么

mmlingyu

1.模型路径问题？ 2.网络连接问题？3.加速器命令行工具错误？我在使用魔法啊，但是连接不上Hunggingface的模型。

2

The following values were not passed to `accelerate launch` and had defaults used instead: `--num_processes` was set to a value of `1` `--num_machines` was set to a value of `1`...

ImCVer