This might be one of the simplest workflows you've seen. Why not give it a try?
Workflow for MOSS Image
T2I + BBox: use the bounding box to control where the character is placed, or ignore it and use T2I as usual.
I2I + Depth: get the same pose as your reference.
Hires: adds a second pass to refine details and make the image sharper.
Note : Update ComfyUI to the latest version and enable Node 2.0.
How to Use
T2I & BBOX+Hires
Describe what's there and here, boom. Too lazy? Ignore the BBox and just write what you want.
I2I + Depth
Convert any image to a depth map, use the I2I depth workflow, write your prompt, done. Increase the strength if you want the result to follow the reference more closely.
Using a LoRA
There's a LoRA node inside the subgraph. To access it, click "Enter subgraph" on the MOSS-Image node, then enable it and plug your LoRA in.
The model link is already included in the workflow note. If you can't find it, get the MOSS model from the official ComfyUI page here
If you run into any errors or issues, leave a comment!
Description
- Add new Hires fix pass stage 2, it adds more detail and gives you free upscale.
- Lightweight custom node to fix the BBOX aspect ratio.
- Added character_boxes input.
- Simple prompt copy logic from stage 1 (leave blank if you don't want to change it).
- Changed recommended VAE to flux2-vae_small.
- Added links for 4-steps and 2-steps Turbo LoRA alternatives.
- Uses native nodes, so you don't need to download extra node packs.
- Enable node 2.0 if you want to use BBOX feature.
- As always, I made it simple so don't think too much about it, just download and run :)