This is a spin on the original Kijai's Wan 2.2 Animate Workflow to make it more accessible to low VRAM GPU Cards. ⚠ If in doubt or OOM errors, read the comments inside the yellow boxes in the workflow ⚠ ⚠ This WF uses SageAttention. If you don't have or need Triton/Torch/SageAttention you'll have to disconnect the "WanVideo Torch Compile Settings" in the "Model Loader" group or else you'll have a Triton related error. ⚠ ❕❕ Tested with 12GB VRAM / 32GB RAM ❕❕ I was able to generate 113 Frames @ 640p with this setup (10min) ❕❕ Use the Download button at the top right in this page 🟣 All important nodes are colored PurpleMain differences:<ul><li>VAE precision set to fp16 instead of fp32</li><li>FP8 Scaled Text Encoder instead of FP16 (If you prefer the FP16 just copy from the Kijai's original wf node and replace my prompt setup)</li><li>Video and Image resolutions are calculated automatically</li><li>Fast Enable/Disable functions (Masking, Face Tracking, etc.)</li><li>Easy Frame Window Size setting</li></ul>I tried to organize everything without hiding anything, this way it should be better for newcomers to understand the workflow process. <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/6fc44a78-f3ed-443a-8c3f-3e897023ba94/width=525/6fc44a78-f3ed-443a-8c3f-3e897023ba94.jpeg" />

Wan 2.2 Animate Workflow for low VRAM GPU Cards

This is a spin on the original Kijai's Wan 2.2 Animate Workflow to make it more accessible to low VRAM GPU Cards. ⚠ Tested with 12GB VRAM / 32GB RAM ⚠ ⚠ Read the comments inside the yellow boxes in the workflow ⚠ 🟣 All important nodes are colored Purple 🟣

Files with SHA256: 7252bf7547e69ae908a234d409d745200d37fb5660a64d19781adf3b8ad5fa16