WAN 2.2 I2V GGUF Workflow for 12GB VRAM v1.0

Image-to-video workflow for WAN 2.2 I2V (GGUF), built to run on a 12GB VRAM GPU.

Workflow

1. Generate an image with Stable Diffusion (Illustrious checkpoint)

2. Turn it into a video with this ComfyUI workflow

Required models

- Video model: DaSiWa-WAN 2.2 I2V 14B TastySin v8 | Lightspeed | GGUF (High and Low)

I use Q6 High and Q6 Low on 12GB VRAM. Choose the GGUF size that fits your GPU.

- LoRA: SmoothMix Animations WAN 2.2 (High and Low)

- Custom node for loading GGUF models (e.g. ComfyUI-GGUF)

Settings

- Resolution: 768 x 768

- Length: 81 frames / Frame rate: 32

- Steps: 5 total (3 on High, 2 on Low)

- Sampler: lcm / Scheduler: beta

- CFG: High 2.4 / Low 1.0 (adjust depending on how well the prompt works; above 3 tends to break the video)

- LoRA strength: High 0.7 / Low 1.0

Tips

- If the video barely moves from the original image, lower the LoRA strength on the High model.

- If a specific motion looks off, add a LoRA made for that motion.

- One video takes about 15 minutes on my 12GB GPU.

Description