Added v2.0
Updated the workflow to support the latest nightly build of the Wan22FMLF node (commit 7140cd2). In my testing, the node now improves the momentum carried over from the previous latent, resulting in much smoother transitions now. You should update the node to use v2 workflow.
__________________________________________________________________________________________________
Added v1.0.1
Display stitched result on each segment.
Added single extend version,
It lets you extend the video infinitely (one segment at a time) by feeding the last output back in. Max 81 frames or 5 seconds extend
--
I’ve been using Stable Video Infinity (SVI) to extend existing videos for couple of days, and honestly it’s pretty impressive for what it’s trying to do. Instead of generating a whole video from scratch, my workflow approach focuses on taking an already existing clip and continuing it forward in a more coherent way and It can be used with loras and different prompt on each segment. All download links for required Models is on the workflow.
1. Load your video you want to extend
2. Input the steps (minimum 4 steps using lightx2v speed loras) and resolution
3. Input the length (max 81 frames)
4. Input the prompt and loras
5. Profit
Use CFG > 1 and 20 steps if you are not using speed loras,
Euler Simple is fine most of the time
Required Custom Nodes for the workflow, you can install through Comfyui Manager :
ComfyUI-Wan22FMLF (SVI implementation) (nightly) (commit 7140cd2) : https://github.com/wallen0322/ComfyUI-Wan22FMLF
ComfyUI-GGUF (if you're not using FP16/FP8 model) : https://github.com/city96/ComfyUI-GGUF
ComfyUI-VideoHelperSuite (v1.7.9): https://github.com/Kosinkadink/ComfyUI-VideoHelperSuite
ComfyUI-KJNodes (nightly v1.2.3) : https://github.com/kijai/ComfyUI-KJNodes
ComfyUI-Easy-Use (nightly v1.3.5): https://github.com/yolain/ComfyUI-Easy-Use
ComfyUI-MediaMixer (nightly) : https://github.com/DoctorDiffusion/ComfyUI-MediaMixer
rgthree-comfy (1.0.2512112053) : https://github.com/rgthree/rgthree-comfy
Optional Custom Node :
mikey_nodes (1.0.5) (organize output folder) : https://github.com/bash-j/mikey_nodes
Description
Added single extend version, It lets you extend the video infinitely (one segment at a time) by feeding the last output back in. Max 81 frames or 5 seconds extend.
FAQ
Comments (18)
Simple and powerful, thanks
Got this error in both 1.01 and single extend
Prompt outputs failed validation: WanAdvancedI2V: - Value 0.0 smaller than min of 1.0: structural_repulsion_boost
okay, change structural_repulsion_boost to 1 and it starts but I'm not sure if this is correct
hm, with these settings, when creating a new clip, it starts twice in the already merged one and with a fade effect
having the same error, where did you change this value? cant find it..
@hook0r inside ksampler
@fullofbenzo it doesnt work with 1.0. whats the right value?
@fullofbenzo I have the exact same issue (fade effect).
Did you find a fix ?
@BretChampagne nope
@fullofbenzo
@fronyax
Ok, thanks anyway for this. I Guess I ll still play with this workflow desactivating the "stitched" part...
@fronyax
Just don't understand how it can work for you with this "structural_repulsion_boost" issue...
@fronyax
Thanks anyway, love the idea to send in a 1rst clip to get such accurate 2nd & 3rd clips + the possibility to fine tune the prompt/loras each time...
I m just a bit sad it doesn t work as expected when it comes to blend the whole thing...
@fullofbenzo have you found another SVI workflow that worked for you ?
I loved the "promise" of this one because of the vid 1rst input, the overall flexibility (and apparent simplicity)
I have no luck with these fucking SVI workflows untill now... I guess I m dumb...
@BretChampagne I'm sorry it doesn’t work for you, I wish I could help more. I haven’t been able to replicate the issue you experienced on my side. The only fade issue I’ve run into so far happens when I accidentaly use a 5 second input video that was "trimmed" (skip first frames) to about 3 seconds by the VHS Load Video node. In that case, the output tends to fade and doesn’t connect properly to the next clip.
@fronyax Thanks for your answer. You don t have this "structural_repulsion_boost" error when you try to make a render just after dropping your json in comfy?
Like @fullofbenzo, It won t work for me if I don t set it on 1 (and not 0)... Then the workflow produce good quality clips, but not continuous.
@BretChampagne I have that once when i was trying the node for the first time, but after reinstalling, completely delete the folder, and updating it, the error was gone and can generate normally.
I got this running and it works great but it completely changes the faces of the people in the clip and re-skins the entire image, basically... what could I be doing wrong?
THE FACES VARY ON EACH ITERATION WITHIN WAN.. YOU'D HAVE TO FACE MATCH THE PERSON ON THE LAST IMAGE TO KEEP CONSISTENCY OR MAKE A LORA FOR THAT PERSON