WAN 2.2 14B Low-VRAM Text/Image-to-Video
Run state-of-the-art video generation locally on GPUs with as little as 6GB VRAM. WAN 2.2 14B brings top-tier text-to-video and image-to-video quality in a format optimized for low-VRAM devices. Quantized GGUF builds keep output high-quality on consumer GPUs, with FP16/FP8 variants available for more powerful hardware.
✓ 100% free — download the workflow and run it locally.Requirements
- GPU: NVIDIA RTX 30XX / 40XX / 50XX (FP16 support required)
- VRAM: 6 GB minimum
- OS: Windows
- Storage: 40 GB+ free
Model downloads
- umt5-xxl text encoder (Q5_K_M GGUF) https://huggingface.co/city96/umt5-xxl-encoder-gguf/tree/main
- wan_2.1_vae.safetensors https://huggingface.co/Kijai/WanVideo_comfy/tree/main
- Wan2.2 T2V A14B LowNoise (Q3_K_S GGUF) https://huggingface.co/QuantStack/Wan2.2-T2V-A14B-GGUF/tree/main/LowNoise
- 2xLexicaRRDBNet_Sharp upscale model https://huggingface.co/Thelocallab/2xLexicaRRDBNet_Sharp/blob/main/2xLexicaRRDBNet_Sharp.pth
- WAN 2.1 T2V 14B lightx2v LoRA https://huggingface.co/Thelocallab/WAN-2.1-loras/tree/main
- Standard WAN 2.2 diffusion models (FP16/FP8) https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/tree/main/split_files/diffusion_models
Custom nodes
- ComfyUI Manager https://github.com/ltdrdata/ComfyUI-Manager
- ComfyUI_LayerStyle https://github.com/chflame163/ComfyUI_LayerStyle
- ComfyUI-KJNodes https://github.com/kijai/ComfyUI-KJNodes
- rgthree-comfy https://github.com/rgthree/rgthree-comfy
- ComfyUI-GGUF https://github.com/city96/ComfyUI-GGUF
- gguf https://github.com/calcuis/gguf
Setup steps
- Download and sort all models into their ComfyUI folders (clip, vae, unet, upscale_models, loras, diffusion_models).
- Install the custom nodes above via ComfyUI Manager.
- Load the workflow JSON below into ComfyUI.
- Enter a detailed text prompt (enhancing it with an LLM gives the best results).
- Use the Fast Groups Bypasser levers to enable or disable workflow sections.
- Generate — expect ~10–15 min for 480p on an RTX 4050, faster on higher-end GPUs.
Notes & tips
- Choose between GGUF or standard diffusion models directly in the workflow.
- Detailed prompt engineering is recommended for the best output.
- One-click Windows installer available if you'd rather skip the manual setup.
More resources
- RunPod cloud template https://get.runpod.io/WAN2-2-Video-Template
Want the 1-click version?
Skip the manual model downloads and node installs — get a ready-to-run installer, or unlock all 75+ with Local Lab Pro.