There's already plenty of workflows on this damn site, but I wanted to share the ones I've personally made for myself.. why not, I learn from the community and give back to it.
They're relatively simple, in my opinion. I don't like LLM crap, overlapping nodes, over complication or anything that makes a workflow look like it was inspired by a setup in Satisfactory or Factorio.
The second one is a bit more updated. If you can improve my workflows without over complicating it, feedback is welcome.
Custom Nodes
Acquire these from the Comfy Manager, or install them via their github link. I use Nodes 2.0 so keep that in mind.
I also use rgthree in general for their Fast Toggle Group Headers setting, but nothing from the pack itself.
You are expected to have the video and audio models for Minimax, but you will likely get errors from these two. I.. wouldn't recommend downloading these two files (45GBs in total) just to get it to work. I have over 150 GBs of stuff downloaded for giggles.
Download/use your own and switch them out.
Main Workflow
Supports one video reference + up to 4 image references (bypass what you don't use)
Features a basic prompt switch, you can save up to 4 prompts but it only does one at a time.
Use "select" on the Prompt Switch to select the Prompt Box you're using.
I load images at full resolution, generate at a lower resolution, and then upscale back to original resolution
Performance Nodes enabled by default.
Sequence Workflow
Features 4-single image reference with their own prompt box, with their own length switch but they're connected to the same LoRAs
You can use it to run four jobs in sequence. The same four images with different prompts, or four different images, different prompts and different duration.
Sequence controls where the first job starts and batch count controls where it stops. Sequence 1 = Image A, and batch count 4 means it does all four. See markdown notes if needed.
I won't be covering system specs, recommendations and the sort.





