VERSION 3 NOTES
First off, this has a custom node I made for predicting flow to help smooth transitions iteration to iteration. Place the whole folder (included in workflow download) in your custom nodes.
Version 3 is an experiment at this point. Could use some more testing and tweaking, but it has some cool features so wanted to share.
Predictive Flow - this is probably the coolest one and is working the best right now. Basically at the end of an iteration, it predicts the beginning of what the next iteration should start like. This gets converted to a latent to get blended in with your subsequent iteration. Should reduce jitters from iteration to iteration
Noise adaptation - if your iterations start losing quality and get noisier, it should dynamically switch to adding more steps and adjusting sampler parameters
Face ID - Not working yet since I'm running into wan compatibility issues, but it's intended to prevent face warping. I'm working on this still.
So from version 2, I added a new T2V feature for it. With this, you can change the workflow to start from a T2V prompt creation instead of an I2V workflow for the first iteration. Then subsequent iterations will go off I2V so you can continue your T2V prompt for however many loops you want.
Heads up that FLF on T2V still references the uploaded image. I'm wanting a future version to reference the first image of the T2V created vod, but I haven't got the conditional logic set up yet.
Notes on credits:
I got the base of it from https://civarchive.com/models/1829052?modelVersionId=2070152
I got most of the florence stuff from https://civarchive.com/models/1687498/wan-2221-i2v-2-workflows-merge-fusionx-lora-2-sampler-florence-caption-last-frame-color-match?modelVersionId=2061133
The stuff around FLF was mainly me messing around so I added that myself and also made some adjustments from the previous two workflows mentioned to what I liked more.
What does it do?
It's basically a for-loop gguf wan2.2 workflow. You can set up however many iterations to go through. There is an option for the last iteration you have to go to a FLF workflow which should help continuity a bit.
Features:
For-loop based I2V gguf workflow
T2V first iteration to I2V gguf workflow
Auto captioning per iteration + customizable after text per iteration
Auto sizing images (to avoid mat errors)
Last For-loop iteration goes to FLF (optional selection)
Upscale + Interpolate
Individual Loras per high and low lora per iteration (power lora so easier selection)
Using ClownsharKSampler
To be added at some point... maybe (once I study up on these more)
VACE integration (wanting to look into Phantom too)
Spline Integration
So far I've been finding good results with linear/euler + beta57, but always looking for better options.
Description
FAQ
Comments (18)
Thank you, but I have simple rule for WF. To never download an unorganized, messy workflows :-) They can work, but are pain to use or change.
It is not worth the time messing with it. But I'm sure someone will find it useful.
At least take a look at it before leaving a comment to call it useless 😂. But oh well no worries. There are plenty of workflows out there, I just wanted to have something better than pastebin for sharing.
I’m not calling WF useless, this is misunderstood. Just wanted to point out that WF should be nice and easy to understand. Definitely should not be overdone or hard to follow.
vladulidlo and my doctor says I should work out and eat fewer cookies but what kind of life is that?
Why comment just to say "it's not good enough for me?" Are you some celebrity that people are going to care about your opinion on this sort of thing? Did you think gumpbubba uploaded it here for you? Or do you expect gumpbubba to hear your advice and change how they work, never having previously been exposed to the radically new concept of making something "nice and easy?"
gumpbubba721291 Thanks, even if it's messy it's still a good example of a few things I haven't seen done before so I might learn something.
nooooooooooooooo It is just my opinion and nothing more :-)) There is no need to make great deal out of it. Just stating my opinion, is it alowed?
I don't agree with this, for the complexity of the WF is pretty well sorted imho.
Nobody here understood the hyperbole. It was not meant to criticise this particular WF. It was about the statement of the author about WF being unorganised and messy, which is not even true for this particular WF :-)
Although it's best to download a WF and then say it's messy after seeing it - For what it's worth, when the original comment was made, it was on the v1.0 which was A LOT messier than the current version. I cleaned things up a lot in 1.1 and v2
gumpbubba721291 I was commenting on the statement you made, not the WF itself. The statement worked like a red flag for me. I actually just finished repairing my ComfyUI (again) after trying WF from different autor with specific custom nodes that installed gazillion Python libraries and broke my ComfyUI. My comment was not meant to do harm to your work, it was the other way.
vladulidlo No harm taken, glad you enjoy the latest version 🙏
The process hangs on "Auto Caption Stuff" for a long time. Is there any way to speed up the process?
The way it is now, it will take a long while to load the first time, but once loaded, it will be cached and be fast additional runs. However, to reduce time for the first time as well, I would consider changing the model Florence-2-large to something else. I'll have to double check on the time differences between each later and get back to you on it
gumpbubba721291 I changed it to GGUF Q4_K_M and everything became lightning fast!
Awesome good to hear
Thanks, I've already improved it. Just a question: I read somewhere that in some cases it is not the best thing to do to enable Torch compile and/or Sage Attention, because it causes problems. Is that why it is not enabled here? (I don't remember where I read it) or again i missed some note or node ^^
Awesome. I've also heard of issues with sage attention hitting quality. I haven't heard of the same issues with torch as sage attention. Torch compile would probably be a good thing to add.
gumpbubba721291 ok I try. because without, it's very slow here, even with a rtx 5090
argotic Valid. In full disclosure, I'm running this on a H100, so I'm generally trying to get the highest quality and not too concerned around speed because most things are going to be fast then.