This contains subgraph features. If you have node issues, please update the ComfyUI frontend.
Most of my videos contain workflow. (But maybe CIVITAI re-encoded it so you won't get it.)
This is the basic WAN2.2 14B T2V, I2V workflow. The custom node only uses gguf.
By default, this creates a very low-quality 480P preview.
(HIGH model only, 2step)
If you like it, turn on the UPS group, do a V2V at 720P, and save it in high quality.
(LOW model only, 2step)
Faster than 720p 2/2 step, but may be unstable,
but it's nice to be able to preview it sooner.
nothing special.
Description
FAQ
Comments (12)
There is an unknown node (I suspect it's the UPS group that the notes refer to) but it's not recognized by manager.
5147e05e-4f1e-4f7a-b57a-d2cc2a0c5631
I'm sorry my bad.
I guess I just copied my file without exporting, which caused the error.
It should work now. I also included an ungroup version in case it was a subgraph bug.
But it's really nothing special, and the results aren't great.
Fantastic workflow - thank you!
Your work looks amazing.
I tried updating several nodes, but there are two nodes which i am told is missing.
30ea81ea-bdad-437a-9ea3-6dd759237dbb
8be4e531-7d5e-452e-b7af-b78b20f8202e
I am starting with video so i am a little new, but so far i did thought i had everything installed, creating pictures are working out very well.
Can you tell me what these missing nodes are? And how to solve it?
PS: There are two red squares around the Menu Left and Right of the Ksampler (Advanced) menu all the way down on the I2V json chart.
Try updating comfyui.
(I only use the portable version, so I don't know about other versions, but run update_comfyui_stable.bat in the update folder.)
Or Redownload the workflow and try the ungroup version.
(I reuploaded the file because some nodes in the previous file weren't ungrouped.)
@m8rr
Thank you for the tip, i was able to get the workflow to open after i updated it from the folder like you mentioned.
I downloaded the required models to get it to run.
The loaders do not work with any other models at all, and i have no idea what is in the models i had to download for this workflow.
I finally got it to run, but then an error popped up.
KSamplerAdvanced Cannot find Triton??
I worked with different programs and i do not recognize this.
Have you any idea what your workflow means?
I hope to get it running.
@longjohnmonster15inch492
Try bypassing or removing the torchcompile node. torchcompile makes the process a bit faster, but it's not necessary.
@m8rr
I got it to work after i added the right diffusion loader that matches the specs.
Its generating, and its quite decent at that.
But the video does not match the input of words.
It finishes, but sends me this error in the promt,
This explains why it differs from what i entered.
Error lora diffusion_model.blocks invalid for input of size
Have you experienced this before? and do you have an solution?
@longjohnmonster15inch492
I'm not sure. Did you add another LoRa?
If you check T2V and I2V and check HIGH and LOW, there's no reason for it to break.
Download my latest video, drag it into ComfyUI, and run it. You should get a similar video.
Thank you very much for all of your help, i wanted to let you know the schedule helped me build a very basic screen which helped me, fine tuned it. Decent vids are coming out. I wouldnt have thought this possible. even LORA models are now loaded which are helping i hope to find more,, i have heard of different video AI models and fine tuning which i hope to look into.
Maybe a stupid question, how do I just run the upscalers on my low-res preview vid without running the initial i2v gen again?
If you keep the seed value of the last queue, the result is preserved and not regenerated.