Feedback
freeTrialImage.bannerPity
freeTrialImage.upgradeUnlock
- ✓freeTrialImage.benefitHd
- ✓freeTrialImage.benefitWatermark
- ✓freeTrialImage.benefitUnlimited
MOSS-Image
Type a prompt or a JSON layout and MOSS-Image renders objects, text and boxes exactly where you place them — free to try online.
AI Tools
Explore AI tools
Free AI Image
Truly Free AI Image Generator
AI Image Editor
AI Photo Editor Online
Image Upscaler
Enhance image resolution effortlessly
AI Color Changer
Transform photo colors in seconds
AI Room Designer
Design your dream room with AI-powered 3D rendering
Skin Enhancer
Enhance skin texture naturally with AI
Face Editor
Enhance face texture naturally with AI
AI Clothes Changer
Inside MOSS-Image: Layout-First Image Synthesis
A 5B text-to-image family where the prompt itself dictates the layout, not guesswork.
- Layout Control Baked InAssign a box to every object and text string, so the composition follows your JSON instead of the model's assumptions.
- 5B Backbone with 4K-Token PromptsLong, detailed prompts fit without trimming, and the same weights can be fine-tuned for your own tasks and domains.
- Base and SFT CheckpointsBoth checkpoints ship under Apache-2.0 with an identical architecture; the SFT build keeps whiteboard text noticeably tidier.
Running MOSS-Image in Three Steps
From setting up the environment to rendering a fully controlled layout.
MOSS-Image: Frequently Asked Questions
Answers on layout prompts, checkpoints, hardware requirements and licensing.
What exactly can I control with native layout prompts?
Each JSON prompt describes the scene, its objects and any rendered text, assigning a bounding box for every element's position and size.
What separates the Base and SFT checkpoints?
Both use the same architecture and sampling interface; SFT is fine-tuned on curated data and, in the project example, returns clearer text grouping and a neater whiteboard layout.
How many tokens can one prompt hold?
Roughly 4K tokens, which is plenty for a multi-element layout written in JSON.
Which hardware and software are required?
Python 3.10+ alongside an NVIDIA GPU and a matching PyTorch build. When memory runs short, drop to Sequential mode or shrink the image; Compile DiT stays off by default and is not available in that mode.
Is there a ComfyUI integration?
Yes — install the native plugin through ComfyUI Manager, then load the bundled example workflow. Merging it into the official ComfyUI repository is still underway.
Which license covers the weights and the code?
Weights ship under Apache-2.0, while the pipeline code and third-party components retain their own licenses. If you use it in research, please cite the project repository.
Put MOSS-Image to Work Today
Set up the Diffusers pipeline or the ComfyUI plugin, test a JSON layout prompt, or shape your idea with the built-in Prompt Enhancer before you generate.