Description

Most AI image generators are good at mood and composition. Ask them for a diner menu with twelve prices, and the food looks great while the prices turn into nonsense. MossImage is built for the other half of the job: the words.

MossImage runs MOSS-Image SFT, the fine-tuned variant of the open-source MOSS-Image model. You write a prompt, quote the exact copy you need, and optionally switch on Layout JSON to pin text and objects to precise positions. It handles long English and Chinese copy, dense menu grids, packaging labels, ad headlines and UI microcopy — the cases where generic image models usually break.

Four render tiers cover the workflow: Draft at 512 px for spelling and composition checks, Final at about 1 MP for web and social, HD at about 2 MP for print and large displays, and Max at up to 4 MP with a 2048 px side cap for dense layouts. Five ratios are supported: 1:1, 4:3, 3:4, 16:9 and 9:16. Everything runs in the browser; there is nothing to install and no GPU required.

Pricing is credit-based. Signed-in users get 2 free credits per UTC day. Subscriptions start at $9.99/month or $95.88/year, and one-time packs start at $4.99 for 15 credits. Pack credits never expire, subscriptions can be canceled anytime, and technical failures within our control return credits automatically. Paid plans and packs unlock watermark-free exports; commercial use is permitted under the Terms.

The MOSS-Image generator is text-to-image plus JSON layout — there is no image-to-image or inpainting in this workflow yet. The model weights are Apache-2.0 and a ComfyUI workflow is available if you want to run it locally.