-
start websocket server
Connection state is connecting
Connection state is connected
[libx264 @ 0x7f55f8001740] using cpu capabilities: MMX2 SSE2Fast SSSE3 SSE4.2 AVX FMA3 BMI2 AVX2
[libx264 @ 0x7f…
-
**i try serveral times to settle the envirement problem,and still fail in loading the node with the message as follows:**
[ComfyUI-Manager] default cache updated: https://gitcode.net/ranting8323/Co…
-
![image](https://github.com/TMElyralab/MuseTalk/assets/169735487/fa344b9d-e575-46ce-9d7e-a0b99eafa1d6)
-
大佬,我直接用的autodl上的镜像运行,没有显示视频。是哪里不对吗
![image](https://github.com/lipku/metahuman-stream/assets/31032620/058dc4eb-a586-49a6-b3e0-fe64edf8538a)
![image](https://github.com/lipku/metahuman-stream/ass…
-
ComfyUI-MuseTalk这个插件会直接加载模型进GPU,占用显存,一启动comfyui就占用了7.5显存,请看看是否有这个问题
如果是的话,是否可以把模型剪切出来,就不会加载节点,然后就没有东西占用显存了。当然具体原理我也不是特别明白,还请排查一下
-
1. 使用chatgpt改编了延边刺客的经典台词。
2. chatTTS生成语音。我用到了https://ixo96gpj4vq.feishu.cn/wiki/SjiZwNSuUin9fukZh23cIMBOnge 分享的电台主持人语音,并把转成csv在https://github.com/jianchang512/ChatTTS-ui使用。
3. 用MuseTalk生成口型:https://…
-
No CUDA runtime is found, using CUDA_HOME='C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4'
ERNeRF导入失败,原因: list index out of range
使用ERNeRF前需要安装对应的环境
MuseTalk导入失败,原因: DLL load failed while…
-
运行后可以打开页面,输入文字后tts成功,随后就Connection state is failed和Connection state is closed了
flask_sockets.py已经改成websocket=True
(/root/autodl-tmp/nerfstream) root@autodl-container-22da11ad52-3d284594:~/autodl-t…
-
你好,在我们的的部署中发现一个问题,即在VAE的模型中,将GPU上的数据拷贝到CPU上花费了巨量的时间。简单来说就是在不考虑这一步的情况下,实时性可以达到60+的fps。但是因为它的存在导致我们的性能只能在30fps左右。请问有没有什么办法在这个基础上做到优化呢?这是因为显卡位宽所导致的吗?我们的实验环境是4090。
-
```
please download ffmpeg-static and export to FFMPEG_PATH.
For example: export FFMPEG_PATH=/musetalk/ffmpeg-4.4-amd64-static
Loads checkpoint by local backend from path: ./models/dwpose/dw-ll_uco…