Troubleshooting Wan2.2 in ComfyUI: Channel Mismatch Errors
lazarus102 · reddit · 2026-08-07
A user encountered technical roadblocks when running the Wan2.2 model for image-to-video tasks in ComfyUI. A persistent error occurs at the Ksampler step, indicating an input channel mismatch (expected 36 channels, but got 32).
The user tried updating ComfyUI, swapping various recommended VAEs, and testing different model version combinations, but the issue persists. Interestingly, accidentally using Wan2.1 models bypassed this specific error (though it threw an OOM error instead). This suggests Wan2.2 might have specific compatibility requirements for workflow configurations.
More from Multimodal
- 15-Person Team Creates $500K AI-Generated Cannes Film, Open-Sources Assets — FinanceYF5 · 2026-08-07
- 15-Person Team Uses AI to Create Cannes-Bound Film, Higgsfield Open-Sources All Assets — littmath · 2026-08-07
- Projectify: Seamlessly Run ComfyUI Workflows Directly Inside Blender — AccordingInspector58 · 2026-08-07
- AI Podcast Recap: MiniMax H3 Video Model and Cloud Tooling — altryne · 2026-08-07
- Tencent Hunyuan's WorldClaw: agentic 3D open-world generation with editable assets — Tencent-Hunyuan · 2026-08-07
- Seedream 5.0 Pro Layerize Splits Images into Editable Layers via Text — aziz4ai · 2026-08-07