Hello there and thanks for checking out this workflow!
—Purpose—
This workflow is exploring some of the immense potential of CogVideoX models to make creations come to life!
As usual the workflow is accompanied by many notes explaining nodes used and their settings, personal recommendations and observations.
—Features—
Presets to switch on/off the parts of the workflow that matter
Text-To-Video
Image-To-Video
Interpolation-To-Video
Video-To-Video
— Transition Smoothing
— End-To-Start Looping
Post-Processing
— Frame Interpolation
— Upscaling
—Custom Nodes used—
All of which can be installed through the ComfyUI-Manager
If you encounter any nodes showing up red (failing to load), you can install the corresponding custom node packs in most cases through the 'Install Missing Custom Nodes' tab on the ComfyUI Manager as well.
► In case of "execution-inversion-demo-comfyui" you need to set the "Channel" setting to "Channel: dev". (Found in the top left of the ComfyUI Manager Menu)
→ Do not forget to turn the channel setting back to default again afterwards
► If the "Channel: dev" nodes refuse to install due to some "security level" issue, you can modify the "config.ini" file located in ComfyUI/custom_nodes/ComfyUI-Manager and edit that with any text editor.
The entry at the very bottom "security_level" is likely set to "= normal", which is the default and in some cases prevents installation of custom nodes, especially when using online services.
When you change that setting to "= weak", save and restart Comfy, you should be able to install the nodes just fine.
Links to required/recommended models, along with the folders they need to be placed in, are included.
—Thanks—
The workflow would not be possible as is without these custom node packs. If you want to show your appreciation to the node creators, drop them a ⭐ on their github repos! Thank you!
Feel free to ask any questions, share improvements or suggestions in the comment section!
Also let me know if you encounter any confusing points I can elaborate on and focus on improving for the next update!
Description
v1 — initial release
FAQ
Comments (25)
Looks good, the pink styling kinda confuses me as it looks very similar to the red of the error nodes. I ran into the following too, all custom nodes are installed using the manager script:
When loading the graph, the following node types were not found:
[object Object]
workflow>Image From Batch +
workflow>Input Select
workflow>Default Width + Height
workflow>Revert Loop Offset III
workflow>Reintegrate Transition
workflow>Transition Loop Last To First II
workflow>Smooth Transition At Index III
Hey there, thank you! Yeah the coloring confused others before 😅
1.) Did you install the execution-inversion-demo-comfyui like this :
► In case of "execution-inversion-demo-comfyui" you need to set the "Channel" setting to "Channel: dev". (Found in the top left of the ComfyUI Manager Menu)
→ Do not forget to turn the channel setting back to default again afterwards
It is not showing up in the "Default" channel's list for some inconvenient reason, so the "Install missing nodes" feature likely misses that too.
2.) Aside from that do you get any (IMPORT FAILED) errors in the console when starting up Comfy?
What is your Cuda + torch version? because some nodes werent designed for the latest cuda and torch, try reverting back cuda 12.1 and torch 2.2.0, some nodes like insightface and others dont work with latest versions.
@Plingus I am running Cuda12.4 with torch 2.4.0 myself and it works just fine for me. I think something might have issues loading/importing properly. The "Image From Batch +" group node for instance uses exclusively nodes from the ComfyUI_essentials pack
I will explain what i encountered, 1st time loading this workflow i was just missing the demo + Smooth Transition At Index III + Input Select. i bypassed the latter 2 and installed the demo and the workflow works fine then i wanted to try the fast mode wich needed the cuda 12.4 with torch 2.4.0 , but just before i realized insightface doesnt work with DPM++ on newer torch (only euler) i reverted back to cuda 12.1 and torch 2.2.0, now this recompiles alot of stuff and somehow when i reloaded this workflow all nodes magically where avaiable.
@Plingus That is odd, all these cross dependencies resulting in everyone's installation of comfy being unique is really an issue. I hope the standalone desktop version they are currently testing will allow normalizing those issues down to one working environment everyone can use without even thinking twice about custom node pack requirements.
Just today I went through the trouble of trying to get sageattention to work, as I do already have triton working. I also upgraded torch to 2.5.0 in the process and ran into errors so cryptic I will not even bother figuring them out.
Just fyi with cuda 12.4 and torch 2.5.0 I can run FaceID IPAdapter workflows with dpm++ samplers without issue using insightface. At least all looks normal from the quick test I just did.
@redpinkretro i know this odd, but after researching this error further i came across a error that relies strongly on the installed cuda packages. after recompiling those i had no errors and the missing nodes were all healthy.
@Plingus That is all that matters at the end of the day 👍
Im a bit lost on how to get those nodes all set. says i cant get the demo one with my security level?
Hey there, that does sound odd. So you did set the channel to "Dev" I presume in order for it to even show up in the list, but then it refuses to install still?
Or did you try the "Install via Git URL" option? Cause that one always gave me some security level warning as well for whatever reason.
Use your filebrowser to navigate to ComfyUI/custom_nodes/ComfyUI-Manager and find the config.ini , use a text editor to change the line to security_level = weak and save it, then restart comfy.
@Plingus Very good to know, thank you! 👍
so i had to change the ini file regardless of dev channel setting for comfy manager or git fyi. buuuut those werent the missing nodes it looks like and i swear i have all the others already. in pre-processing A and B and transition smoothing ive got the missing nodes. ones smooth transition at index?
@zimzaderk_ Interesting that that is possible as well, good to know. But the fact that it still refuses to run is not ideal. The groups [Pre-Processing A + B] only use the following node packs :
ComfyUI-Impact-Pack
execution-inversion-demo-comfyui
So I would check next if one of those is not up to date or failing to import. You can also send me your Comfy start-up console log, so I can look for errors there.
You can just https://wormhole.app/ a .txt file of the log
The "Smooth Transition At Index" group node is bit larger and contains all of the above except rgthree, and additionally the following :
ComfyUI-VideoHelperSuite
@zimzaderk_ Had this exact issue. Restarting Comfy, refreshing the tab, closing tab and opening a new one, didn't work. Clearing the workflow and reimporting the json fixed the issue. After ensuring everything was installed correctly of course.
@Hopnoggin Yeah, Comfy can basically "remember" the faulty dysfunctional state of nodes even after they have been fixed. #NotSoComfyComfyThings 😄
@Hopnoggin I was reading this comment thread because I had the exact same missing nodes. Re-loading the .json file fixed the issue
@succupon Glad to read that! I am working on an update to simplify the structure greatly by taking the presets apart into respective workflow groups separate from one another. I think the density resulting from trying to reuse as many nodes as possible by all the presets feels a little claustrophobic to use and confusing.
Is this workflow broken?
Tried everything to get this to work.
ComfyUI updated and in dev mode, ini file modified with weak security.
Still not able to install missing nodes.
Says I'm missing:
- CogVideoImageInterpolationEncode
- CogVideoXFunSampler
Hey there, sorry about that. It broke after a recent update improved some of the nodes on made some bigger changes. I just didn't get to updating it yet, though I'm 90% done with it. 🙏
@redpinkretro Hey there, the features you have for Cogvideo are exactly what I'm looking for! Sadly I can't get it working in it's current state either due to the [Following Missing Nodes]
I tried rolling back CogVideoWrapper to 1.0 Legacy, which seems to work for other workflows, but alas it was still missing:
Image from batch
Input select
Default width + height
Loop offset 3
Reintergrate transition
Loop Last to first
Transition at Index 3.
Any advice?
@redpinkretro Here's how the errors look on the new UI https://gyazo.com/8c8218b1a2327b4d56b2ac2b94c79b90
@jasonafex Hey there, really sorry this is still causing issues.
When I tried finishing the rework of the workflow some update made it unusable for me as well and couldn't even run with any attention selection without errors. And considering how well hunyuan video and LTX perform I am reworking it from the ground up.
CogVideo still has the most features with actual start-to-end interpolation, but the performance and output of Hunyuan and LTX are far better, and LTX is even incredibly lightweight on top of that.
About your error and how to make it work right now:
1.) Do you have execution-inversion-demo-comfyui installed?
2.) Do all custom nodes load properly upon startup of ComfyUI? I.e. no (IMPORT FAILED) messages next to any of the packs
3.) Considering how CogVideo nodes are missing too, it looks like you would need to revert to an earlier version of that, too. The most recent one before publishing date here on the model page should be the one used.
@redpinkretro The [Install Missing Custom Nodes] tab doesn't detect anything/is blank. If there are specific roll backs required to get things to work, that may need to be added to the description.
This workflow , this workflow and this workflow install everything successfully, though they use different versions of cogvideo and I have to update it every time I want to use a feature that is seperated between them. Having frist/last frame and interpolation on the same workflow would save a world of hassle.
If hunyuan is flat out better as you said, I'll try installing those workflows and see if I get back better results!
@jasonafex For the execution-inversion-demo-comfyui pack you need to switch the channel to "Dev", in order for it to show up in the custom node search. Can you confirm that all the necessary nodes, listed under "Custom Nodes", are installed and load properly?
Hunyuan outperforms CogVideo imo, as it runs a lot faster, has more flexibility in terms of frame count and aspect ratio, as well as a far better prompt-adherence.
However Hunyuan does not have a First/Last frame interpolation yet unfortunately.
CogVideo was working good with some prompts and bad on others, but took so long that it was rather painful to find out for me. But I didn't experiment as much with it due to that, so ymmv
@LatentDream has experimented a lot with CogVideo and would probably be able to answer this much better. He has a workflow that was updated far more recently than mine, too, so it's more likely to still function 😄