Cool idea, but no point to try on Mac

[ShahzodBuriev]

I run the avatarify but its too slow, like it is not even performing anything(macbook pro 2018 13). Thank you for the idea, really cool!

[alievk]

Thank you @ShahzodBuriev There is one important point for us that people keep trying to run on Mac: it shows interest to the technology and motivates us to optimize run-time for Mac and CPU in general.

[vineethsoma]

Pretty cool idea! Just tried this on Macbook Pro as well. Is the reason for the low framerate because the runtime is not using the GPU? I am trying to figure out how to improve the performance.

I am going to try installing CUDA and see how it goes. I never ran such workloads on my machine before. :)

[vineethsoma]

So I just learned that Nvidia CUDA divers are only supported till Mac OS 10.13. It is no longer with Mojave release - https://gizmodo.com/apple-and-nvidia-are-over-1840015246

sigh

In case you are using 10.13 instructions here

[DrexelRose]

I also got this working on my 2015 iMac, frame rate is painfully slow. Very very exciting stuff though, I had previously been playing with Fomm on google colab and was excited to see this implementation of it.

[ambanmba]

Needs to use OpenCL / Metal on a Mac otherwise there is no way the CPU can do the math on its own without the GPU.

[jogibear9988]

I also tried on MacBook Pro 2017. Performance much to low

[alexey]

Same one MacBook Pro 2019, Catalina 10.15 FPS is 0.3

[aviomaksim]

Same on MacBook Pro 2018, Catalina 10.15 FPS is 0.3

[alievk]

Now you can run on Mac with a good speed using Google Colab