MOSS-Image Studio: Compose Images from JSON Layouts
freeTrialImage.bannerPity

Feedback

Result
Preview

freeTrialImage.bannerPity

freeTrialImage.upgradeUnlock

  • ✓freeTrialImage.benefitHd
  • ✓freeTrialImage.benefitWatermark
  • ✓freeTrialImage.benefitUnlimited

MOSS-Image

Sketch your scene or paste a JSON layout, and MOSS-Image places every object, caption and bounding box exactly where you want it — free online.

AI Tools

Explore AI tools

placeholder hero

The Advantage of MOSS-Image

A five-billion-parameter image model whose compositions are written down, not guessed at.

  • Layout You Dictate
    Assign a bounding box to every element and every rendered word, so the final composition follows your JSON spec rather than random placement.
  • Five-Billion-Parameter Core
    Extended prompts stay intact from start to finish, and the released weights can be tuned for your own niche, dataset or domain.
  • Two Checkpoints, One Architecture
    Both variants ship under Apache-2.0 with identical structure; the fine-tuned release keeps on-image lettering cleaner and better grouped.

Running MOSS-Image in Three Steps

From a fresh Python environment to a fully controlled layout render.

FAQ

MOSS-Image: Frequently Asked Questions

Answers on prompts, checkpoints, hardware and licensing for MOSS-Image.

1

Which parts of a picture can I position with layout control?

A JSON prompt describes the scene, its objects and any on-image text, and each element carries its own box for position and size.

2

What separates the Base and SFT checkpoints?

Both share one architecture and sampling interface. SFT has been fine-tuned on curated data and, in the project demo, produces tighter text grouping and a cleaner whiteboard layout.

3

Is there a limit on prompt length?

Roughly 4K tokens — plenty of room for a JSON layout that describes many elements at once.

4

What do I need installed to run it?

Python 3.10+ alongside an NVIDIA GPU and a matching PyTorch build. Short on memory? Switch to Sequential mode or reduce image size; Compile DiT is off by default and disabled in that mode.

5

Does it work as a ComfyUI node?

It does. Install the native plugin through ComfyUI Manager, then open the bundled example workflow; inclusion in the official ComfyUI repository is still underway.

6

How is the project licensed?

Weights ship under Apache-2.0, while the pipeline code and third-party components keep their own terms. For research work, cite the project repository.

Put MOSS-Image to Work

Set up the Diffusers pipeline or the ComfyUI plugin, feed it a JSON layout, and watch your composition land — or shape the idea with the built-in prompt helper first.