WAN 2.1 Self Forcing — Text-to-Video + VACE
WAN 2.1 Self Forcing is an autoregressive video diffusion model that streams 480p video in near real-time — 150–400× lower latency than prior models while keeping quality. Includes text-to-video and VACE image-to-video workflows for 6GB-VRAM GPUs.
✓ 100% free — download the workflow and run it locally.Preview
Requirements
- GPU: NVIDIA, 6 GB+ VRAM
- Output: 480p, ~10 FPS on RTX 4090, ~0.8s initial latency
- OS: Windows
- Storage: 30 GB+ free
Model downloads
- WAN 2.1 Self Forcing DMD VACE (FP16 + FP8) https://huggingface.co/lym00/Wan2.1-T2V-1.3B-Self-Forcing-VACE-Addon-Experiment/tree/main
- umt5-xxl 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
- self_forcing_dmd.pt https://huggingface.co/gdhe17/Self-Forcing/tree/main/checkpoints
- WAN 2.1 T2V 14B lightx2v LoRA https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan21_T2V_14B_lightx2v_cfg_step_distill_lora_rank32.safetensors
Custom nodes
- ComfyUI Manager https://github.com/ltdrdata/ComfyUI-Manager
- ComfyUI-VideoHelperSuite https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite
- ComfyUI-KJNodes https://github.com/kijai/ComfyUI-KJNodes
- ComfyUI-WanVideoWrapper https://github.com/kijai/ComfyUI-WanVideoWrapper
- ComfyUI-OllamaGemini https://github.com/al-swaiti/ComfyUI-OllamaGemini
- rgthree-comfy https://github.com/rgthree/rgthree-comfy
Setup steps
- Install ComfyUI Manager + its requirements and the custom nodes above.
- Place all models in their ComfyUI folders (diffusion_models, clip, vae, loras).
- Load a workflow below (standard or FP8).
- Use extended, detailed prompts (the model is trained on them) and run.
Notes & tips
- Supports both text-to-video and VACE image-to-video.
- Detailed prompts give the best motion.
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.