Closed AlexUrrutia closed 6 days ago
same problem on M1 Pro
Same -- terrible frame rate
https://github.com/gongzhang/Deep-Live-Cam/blob/main/README.macos.md
try this
I did, same result...
Try playing around with the command line parameters for --execution-provider
and --execution-threads
:
I seem to have better luck with using the CPU than coreML. Also try keep FPS
.
python run.py --execution-provider cpu --execution-threads 4
managed to get my version working
https://github.com/jasonkneen/Deep-Live-Cam.git
faster and now uses the GPU on Apple Silicon
managed to get my version working
https://github.com/jasonkneen/Deep-Live-Cam.git
faster and now uses the GPU on Apple Silicon
Can you post the steps needed to make it work with your forked code? Thanks!
same on M3 PRO 512+36GB RAM
At least you got it working 😬
I just got a blank window on both M1P and M2P MBPs. Using pyenv with 3.10.14.
At least you got it working 😬
I just got a blank window on both M1P and M2P MBPs. Using pyenv with 3.10.14.
Try Python 3.11. It's what I am using. Via MiniConda.
managed to get my version working
https://github.com/jasonkneen/Deep-Live-Cam.git
faster and now uses the GPU on Apple Silicon
It appears your requirements.txt is missing customtkinter and it won't run without it. I tried adding customtkinter==5.2.2 from the original repo to your requirements.txt and it runs but won't load the model in live view. It fails to load the model with an INVALID_PROTOBUF error.
managed to get my version working https://github.com/jasonkneen/Deep-Live-Cam.git faster and now uses the GPU on Apple Silicon
It appears your requirements.txt is missing customtkinter and it won't run without it. I tried adding customtkinter==5.2.2 from the original repo to your requirements.txt and it runs but won't load the model in live view. It fails to load the model with an INVALID_PROTOBUF error.
mine is workingworking. But still a lower rate of GPU.
managed to get my version working
https://github.com/jasonkneen/Deep-Live-Cam.git
faster and now uses the GPU on Apple Silicon
Can't get you version working @jasonkneen
Using the M3 version, I attempted the following:
tensorflow-macos
and tensorflow-metal
. python run.py --execution-provider coreml --video-quality 0 --live-resizable
. The performance improved, but it can be better because I'm not using the power of my graphic card.
Apple MacBook Pro M3, how to improve fps?
https://github.com/user-attachments/assets/0a6ebebb-e208-4266-98ae-005fe7ad6e3c