Type | Workflows |
Stats | 216 0 |
Reviews | (20) |
Published | Apr 25, 2025 |
Base Model | |
Hash | AutoV2 53D33A40B9 |
I dont know if this was broken for anyone else, impact pack switch nodes are borked atm and v2.1 needed them. If you updated your impact pack it was probably broken like it was for me
Bug - if you want to do last frame only, you will need to connect the image to the resize, i forgot to add a new switch for that.
V3 - fixes and upgrades
- removed impact pack switches, they are broken with latests updates
- using comfyroll switches
- changed from fun controls to default wan image to video
- can now use the 14b image to video wan models as well as the INP models
- Added model swap for upscale
- fixed kjnodes resizes to v2
- Probably the best interpolation between 2 images currently
V2.1 - Small changes
- Set defaults that should work for most things
- added latentsync option
- Changed end to account for latentsync.
- Changed seed to same for both samplers
- notes explain more.
- Few small qol changes
V2
- added end frame loader allowing frame interpolation
- can use either first or last frame only or both
- added notes with models and nodes used
- Finished UI
- added switches for all groups
- fixed all bugs from 1.1 hopefully.
- fixed crops
- fixed bypass
Model
https://huggingface.co/alibaba-pai/Wan2.1-Fun-1.3B-InP/tree/main
only need the diffusion_pytorch_model.safetensors
Clip vision/VAE/textencoder
https://blog.comfy.org/p/wan21-video-model-native-support
img2vid 1.1 - Low vram fixes
- Added option to crop start frame image
- Added Purge vram nodes to 6 spots lowering vram needs by a ton.
- i think it uses as little as 6gb on renders now.
- (49% of my 12 for 512 base 81 frames)
- (63% of my 12 for 768 base 81 frames) 68% for 1.5 upscale
- changed order of final groups
- added sound adding to end.
- added switch to upscale to pass through base to end groups if bypassed.
- changed encode to tiled for upscale
Uploaded a hotfix
Fixed 2 bugs, and added rescale after the model upscale at end.
Image to vid with wan fun
https://huggingface.co/alibaba-pai/Wan2.1-Fun-1.3B-InP
Tip: can connect the ref image to the end frame to animate into the image.
Simple wan 2 image to video
Using inp model
-8 step base
-5 step upscale
-teacache
-Skip layer guidance
-lower memory image to video.
V2
- added teacache
- added shift
- added feta enhance
- added upscaler
- added florence image prompting
- added frame interpolation
- added audio
- Changed a few things with latents and info
Wan 2.1 workflow
-text to vid
-gguf support
-1.3b model
-13b models
-image to vid option
-nsfw unfiltered
Image to vid GGuf links below
Models
Comfy Update
https://blog.comfy.org/p/wan21-video-model-native-support
I use the VAE from here and text encoder/clipvision
https://civitai.com/models/1295569?modelVersionId=1463630
Main
https://huggingface.co/Kijai/WanVideo_comfy/tree/main
These vaes dont seem to work with this
I used the models from here
13b gguf TTV
https://huggingface.co/city96/Wan2.1-T2V-14B-gguf/tree/main
13b gguf ITV
https://huggingface.co/city96/Wan2.1-I2V-14B-480P-gguf/tree/main