Wan Animate Ultimate 12GB — Wan Animate 2 native or Wan 2.2 Animate GGUF Q5_K_S, character animation with lip-sync
1
19
1
☕ One coffee = one happy creator: https://buymeacoffee.com/the_frizzy1
Make any character move like your driving video, on one dropdown: Wan Animate 2 native at 12 GB, or Wan 2.2 Animate GGUF for less VRAM. An original-versus-result side by side is included. It replaces the 5-custom-node v1.2.0 Animate workflow.
What you need
✔ Custom nodes — Wan Animate 2 mode: none, core ComfyUI only. GGUF mode: ComfyUI-GGUF only, with SDPose for pose
✔ ComfyUI version — 0.37.4 or newer
✔ VRAM — 12 GB — tested, not estimated. Peak 11.4 GB native, 11.7 GB GGUF. The GGUF path runs on less
✔ System RAM — 32 GB. Peak 17.5 GB native, 24.6 GB GGUF
✔ Tested on — RTX 3060 12 GB · 32 GB DDR4 · ComfyUI 0.37.4 (Sage attention)
✔ Licence — Workflow MIT · Wan: Apache-2.0
Models to download
Click a file name to download it. Put it in the folder shown, inside ComfyUI/models/. Every file name was read from the workflow .json. Download only the mode you run.
Wan Animate 2 mode — 4 files, from Comfy-Org/Wan-Animate-2.
diffusion_models/
File: wan_animate_2_distill_int8_convrot.safetensors — 16.7 GB (12 GB (Wan Animate 2))
Download:
https://huggingface.co/Comfy-Org/Wan-Animate-2/resolve/main/diffusion_models/wan_animate_2_distill_int8_convrot.safetensors
Put it here:
ComfyUI/models/diffusion_models/
text_encoders/
File: umt5_xxl_fp8_e4m3fn_scaled.safetensors — 6.7 GB
Download:
https://huggingface.co/Comfy-Org/Wan-Animate-2/resolve/main/text_encoders/umt5_xxl_fp8_e4m3fn_scaled.safetensors
Put it here:
ComfyUI/models/text_encoders/
clip_vision/
File: clip_vision_h.safetensors — 1.3 GB
Download:
https://huggingface.co/Comfy-Org/Wan-Animate-2/resolve/main/clip_vision/clip_vision_h.safetensors
Put it here:
ComfyUI/models/clip_vision/
vae/
File: Wan2_1_VAE_bf16.safetensors — 0.3 GB (Wan Animate 2)
Download:
https://huggingface.co/Comfy-Org/Wan-Animate-2/resolve/main/vae/Wan2_1_VAE_bf16.safetensors
Put it here:
ComfyUI/models/vae/
Wan 2.2 Animate GGUF mode — pick one GGUF, then the five files under it. The text encoder and clip vision above are needed in both modes.
diffusion_models/
File: Wan2.2-Animate-14B-Q5_K_S.gguf — 12.3 GB (low VRAM (GGUF mode))
Download:
https://huggingface.co/QuantStack/Wan2.2-Animate-14B-GGUF/resolve/main/Wan2.2-Animate-14B-Q5_K_S.gguf
Put it here:
ComfyUI/models/diffusion_models/
File: Wan2.2-Animate-14B-Q8_0.gguf — 18.7 GB
Download:
https://huggingface.co/QuantStack/Wan2.2-Animate-14B-GGUF/resolve/main/Wan2.2-Animate-14B-Q8_0.gguf
Put it here:
ComfyUI/models/diffusion_models/
File: rt_detr_v4-x-hgnet_fp16.safetensors — 0.1 GB (GGUF mode)
Download:
https://huggingface.co/Comfy-Org/SDPose/resolve/main/diffusion_models/rt_detr_v4-x-hgnet_fp16.safetensors
Put it here:
ComfyUI/models/diffusion_models/
loras/
File: lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors — 0.7 GB (GGUF mode)
Download:
https://huggingface.co/Comfy-Org/Wan-Animate-2/resolve/main/loras/lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
Put it here:
ComfyUI/models/loras/
File: wan2.2_animate_14B_relight_lora_bf16.safetensors — 1.4 GB (GGUF mode)
Download:
https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/loras/wan2.2_animate_14B_relight_lora_bf16.safetensors
Put it here:
ComfyUI/models/loras/
vae/
File: wan_2.1_vae.safetensors — 0.2 GB (GGUF mode)
Download:
https://huggingface.co/Comfy-Org/Wan_2.2_ComfyUI_Repackaged/resolve/main/split_files/vae/wan_2.1_vae.safetensors
Put it here:
ComfyUI/models/vae/
checkpoints/
File: sdpose_wholebody_fp16.safetensors — 1.9 GB (GGUF mode)
Download:
https://huggingface.co/Comfy-Org/SDPose/resolve/main/checkpoints/sdpose_wholebody_fp16.safetensors
Put it here:
ComfyUI/models/checkpoints/
Modes
One dropdown, two paths, both taking a driving video plus a character image: Wan Animate 2 (12 GB), all core nodes, and Wan 2.2 Animate GGUF, which needs only ComfyUI-GGUF and uses SDPose for pose. Load The_frizzy1_wan-animate-ultimate_v2.0.0.json, pick the mode, add your inputs, Run.
Second video, the lip-sync fix: https://www.youtube.com/watch?v=aqFdglMbcIU&list=PLRe4JVWoY_AU
Speed on my RTX 3060
Each mode run once through this workflow, 5 s clips. Wall is queue to finished including model loading, sampling is time in the sampler nodes.
Wan Animate 2 (12 GB, 81 frames) — 480x864, 10 min 31 s wall, 9 min 31 s sampling, 11.4 GB peak vram, 17.5 GB peak ram
Wan 2.2 Animate GGUF (Q5_K_S, 81 frames, face video) — 480x864, 7 min 05 s wall, 4 min 51 s sampling, 11.7 GB peak vram, 24.6 GB peak ram
Wan Animate 2 at 81 frames is 10 min 31 s for 5 seconds of video, 9 min 31 s of that in the sampler. The GGUF path is quicker at 7 min 05 s, and it pays for it in system RAM: 24.6 GB peak against 17.5 GB. On a 32 GB PC that is close. Peak VRAM is 11.4 and 11.7 GB on a 12 GB card either way, so it barely fits.
What did not work
Those two rows were measured with v1.0.0, and in v1.0.0 the relight LoRA did not load. So the GGUF row ran without relight. v2.0.0 loads it, which means the GGUF row is not a like-for-like number for this version any more. There is no v2.0.0 measurement yet and I am not going to make one up.
If your models live in sub-folders the loaders show red until you re-pick them once.
Everything above is 5 s clips at 81 frames, 480x864. Longer or bigger is untested here.
Get the workflow
GitHub: https://github.com/Thefrizzy1/The_frizzy1-Workflows/tree/main/workflows/wan2.2/animate-ultimate
Hugging Face mirror: https://huggingface.co/The-frizzy1/The_frizzy1-Workflows/tree/main/workflows/wan2.2/animate-ultimate
More videos on this stuff: https://www.youtube.com/playlist?list=PLRe4JVWoY_AU
