# From Prompt to Poster: A Short Guide to Text Rendering With AI
If you design posters, covers or banners for a bilingual audience, text inside the image is the hard part. This guide looks at how the open MOSS-Image model handles it and how to try it through Moss Image without installing anything. For the live version, visit Moss Image Workflow.
Contents
# JSON layout prompts
Besides plain prompts, MOSS-Image understands JSON layout prompts. You describe each element, such as a headline, a product shot or a background, and where it should sit. The model then places each element where you asked, which is far more predictable than hoping a long sentence is interpreted the way you meant it.
# Base and SFT checkpoints
Two checkpoints are available. Moss Image Base is intended for LoRA training and fine-tuning. MOSS-Image-SFT is the tuned version for direct generation, and it is the one the online site runs. If you plan to train your own style, start from Base; if you just want good pictures, SFT is the practical choice.
# Settings for repeatable results
The recommended workflow uses 50 or more steps, a guidance value between 3.0 and 6.0 and the Euler sampler. Advanced settings let you change steps, guidance, negative prompt and seed, and pick an HD or 2K canvas. Fixing the seed while you change one setting at a time is the easiest way to see what each option really does.
# No GPU, Docker image or local install
Running a 5B model locally means a strong GPU, a working driver stack and usually a Docker image. The browser version skips all of that. You type a prompt, pick your settings and get the image, which makes it a quick way to decide whether the model fits your project before investing in local hardware.
# Pricing and credits
The site is freemium. New accounts get free credits at sign-up and no card is needed. Paid credit plans exist too; for example, the yearly plan works out to about $8.3 per month with 120,000 credits per year, 8 concurrent generations, private generation, no watermark and a commercial license. A launch banner currently advertises 50% off yearly plans.
# Bottom line
Moss Image is worth a try if this matches what you need. Moss Image is an independent site and is not affiliated with the OpenMOSS team; the MOSS-Image model weights are released under Apache-2.0.