How to get readable text in AI images
Most image models were trained to make pictures, not to spell. Ask for a poster that says "Grand Opening" and you get "Grnad Openig", melted letters, or text in the wrong place. Long lines and small print are the worst.
How to fix it
- Use a model that can spell. MOSS-Image reads your prompt with a 9-billion-parameter language model, so it understands exact wording. That is the model behind our Text & Posters tab.
- Put the exact words in double quotes. Write "SUMMER SALE" in quotes, or type it in the "Words to print" box. Quoted text is copied letter for letter instead of being treated as a description.
- Say where the text goes and how it looks. "Big cream serif title across the top", "a small white sans-serif line at the bottom". Position, size, colour and typeface style all help.
- Break lines yourself. Use real line breaks for multi-line text. Without them the model chooses its own line breaks, which may not be the ones you want.
- Keep it short, or give it room. A title and two or three lines works reliably. A paragraph in a small space does not. Split long copy into separate blocks.
- For exact placement, use a layout. In the Layout tab you draw a box for each text block and type the words into it. The model was trained on these layouts, so text lands where you drew it.
Try it
The Text & Posters tab is set up for exactly this, with examples to start from and an AI helper that writes the prompt for you.