← sources

MOSS-Image accepts image inputs!

source: reddit (reddit:StableDiffusion)published 2026-10-20 19:58ingested 2026-10-21 02:38

retained text

I just tested if it can use image prompting, because its Qwen 3.5 9B text encoder should accept multimodal inputs, and it actually worked! This unlocks the possibility of turning MOSS into an edit model, and a whole bunch of other things.

I tested it with the 8-step turbo LoRA. It's not perfect, but neither was krea 2 without an edit LoRA.  
WORKFLOW: [https://pastebin.com/v9bcZftJ](https://pastebin.com/v9bcZftJ)

document fetched 2026-10-21 00:30 · canonical https://reddit.com/r/StableDiffusion/comments/1xb2h1k/mossimage_accepts_image_inputs