← sources

Turn any ComfyUI workflow into a one-click model (MiniMax H3 Video / Image, Qwen 2.1 + 7 more included)

source: reddit (reddit:comfyui)published 2026-11-11 14:35ingested 2026-11-12 02:42

retained text

Last time I posted my AI studio app I said a ComfyUI node was “in progress”. It took way longer than I thought, but it’s out.

The idea is simple: set up a workflow in Comfy once, choose how many refs and which settings it takes, save it, and paste one URL into the app. From then on it's just a model in a clean UI, next to the API models. Pick it, type a prompt, hit generate, and everything lands in one history, local and API.

I also adapted 9 workflows I actually tried and wired my nodes into them, 19 models in total, each with a prompt guide you can also use as a Claude skill:  
\- MiniMax H3 (video, fast, extend, image, image fast)  
\- Qwen 2.1 (plus head swap, body swap and outpaint)  
\- LTX 2.5 (plus extend and upscale)  
\- MOSS (plus pose)  
\- Ideogram 4  
\- FLUX.2 Klein  
\- Krea 2  
\- Z-Image  
\- Wan 2.2

I rewired them myself, so some may not behave exactly like the originals. If something looks off, tell me.

The VRAM numbers in the guides are calculated, not tested on every card, so if your GPU says otherwise please tell me and I’ll fix it.

It’s still early, so something will break. Install from ComfyUI Manager, search “Unplatform AI Provider”.

Node: [https://github.com/jefoce/unplatform-ai-provider-node](https://github.com/jefoce/unplatform-ai-provider-node)  
App: [https://github.com/jefoce/unplatform-ai](https://github.com/jefoce/unplatform-ai)

Roast it in Issues, and tell me which workflow I should add next.

document fetched 2026-11-12 00:43 · canonical https://reddit.com/r/comfyui/comments/1xtjbn9/turn_any_comfyui_workflow_into_a_oneclick_model