A simple effective workflow to keep VRAM usage minimal when generating with WAN 1.3b!
Yes this also works on the rtx 20xx series
Works the same on non-NSFW as well. Does not have to be for NSFW. SFW will work exactly the same.
The 1.3b should work by default on 8GB of VRAM but this workflow gives you more room to generate higher resolution and frame videos without any OOM errors. I have generated up to 129 frames 512x512 on wan 1.3b even without this workflow but I don't think you could go that high on a higher resolution without a workflow to keep VRAM usage minimal. This should eliminate most OOM possibility's when using 8GB of VRAM. You might be able to even inference on lower VRAM than that if you lower the resolution/frame count.
All characters are 18+. This is intended for mature audiences only you and you alone are responsible for what you generate. 18+ ONLY!
Description
FAQ
Comments (3)
Do you have a link for where you got the Wan21Fun13blnpi2vfp18... model from that the flow defaults to using? I haven't had any luck tracking it down.
Generation times using the same models in your flow is great on my 3060, but I'm having issues with Loras and wondering if you've found the same. Using Loras from here, I get a lot of errors in the log window.
ERROR lora diffusion_model.blocks.21.cross_attn.v.weight shape
and a lot more similar to that. Could it be that the Lora isn't compatible with the 13B model?
I cannot find 90f1e-4e183.safetensors for the life of me. Does anyone have a link to it?
Looks like we don't have an active mirror for this file right now.
CivArchive is a community-maintained index — we catalog mirrors that volunteers upload to HuggingFace, torrents, and other public hosts. Looks like no one has uploaded a copy of this file yet.
Some files do get recovered over time through contributions. If you're looking for this one, feel free to ask in Discord, or help preserve it if you have a copy.