Words that belong in the picture.
Explore Chinese and English lettering for posters, labels, and visual stories. Always proofread generated text before sharing.
MOSS-Image Generator
Let your ideas take shape.
From a few words to something wonderful.
Create with text, typography, and a little more control.
Create with text, typography, and a little more control.
Open-source model Chinese & English text JSON layout control
Your imagination, a little wilder.MOSS-Image SFT
0 / 6000
Try an idea
Checking studio availability…
YOUR CANVAS1024 × 1024
✦
A little idea.
Endless possibilities.
Your next creation starts with a few words.
Give your imagination somewhere to grow.
A blank canvas is a beautiful beginning.✳
A STARTING POINT
Three ideas. Infinite directions.Follow a little inspiration.
Concept illustrations, not model outputs. Select a card to try its prompt.
A NEW KIND OF CREATIVE SPACEBig ideas.
Big ideas.
Thoughtful details.
MOSS-Image is an image model from OpenMOSS. This independent studio brings its creative possibilities into a simple place to start.
Meet your new creative toolkitA place for every little detail.
Switch to JSON layout to describe where text and objects should go. A more intentional starting point for your composition.
Room to make it your own.
Start in the browser, then explore official weights and ComfyUI workflows when you want a more hands-on creative process.
FROM THOUGHT TO THING
A small spark is all you need.
Imagine something.
Describe the subject, mood, and light. Add exact words in quotes if your image needs text.
Give it a little direction.
Choose a canvas shape, or use a JSON layout to guide the placement of specific elements.
Make it yours.
Generate, review, and download. Refine your prompt to explore another direction.
KEEP EXPLORING
All guides A little know-how goes a long way.
A few things you might be wondering.
What is MOSS-Image?+
MOSS-Image is a 5B-parameter image-generation model family from OpenMOSS, with Base and SFT checkpoints. You can explore its official code and model cards through our guides. This site is independently operated.
Is the generator free?+
When enabled, this studio offers up to 3 generations per network address per UTC day, subject to a shared daily budget. Availability is shown in the studio. There is no paid plan in this initial version.
Can I control the layout?+
Yes. JSON layout mode accepts a scene description and bounding boxes for text and objects. It validates your structure and canvas bounds before submitting. Generation remains probabilistic, so review the result.
Where do my prompts and images go?+
When you generate, your prompt and settings are sent to fal.ai for processing. Results are hosted by the provider. Our service keeps quota counters and short-lived result references; see the privacy page for details.