Linux + AMD (≤16 GB VRAM): Let's collect working image-to-video/text-to-video workflows & optimizations

[Kave00Man]

Hi everyone,

I'm hoping we can turn this into a useful resource for Linux + AMD users, especially those of us with 16 GB of VRAM or less.

My setup is:

• AMD RX 7800 XT (16 GB)
• Ryzen 7 7700X
• 32 GB RAM
• Linux

For image generation, everything works beautifully. The moment I step into image-to-video or text-to-video, though, I seem to spend more time fighting OOM errors than actually generating anything.

Instead of everyone solving the same problems over and over, I'd love to collect everything that works in one place.

Please share anything that has helped you, including:

• Working ComfyUI workflows
• Low-VRAM templates
• Optimization settings
• Launch flags
• Model recommendations
• Memory-saving tricks
• ROCm-specific tweaks
• Links to GitHub repos, videos, or guides
• Anything else that helped you run video models on limited VRAM

I also have another thread running as a poll where people can vote on which platform they're using (Linux/Windows + AMD/NVIDIA). I'm curious to see what the community actually looks like, so if you haven't already, I'd appreciate it if you could vote there as well.

Hopefully this thread can become a reference for anyone trying to get video generation working on Linux + AMD without needing a 24–48 GB GPU.

Thanks in advance to everyone who contributes!

(And yes... ChatGPT helped write this. I always use ChatGPT to write stuff for me because I'm a lazy ass. 😄)

[exaiyo]

I'm also searching for a place to find working workflows on modest AMD GPUs. Can you share you text-to-image gen workflow? What resolutions does it gen? Any good upscaling solution?