Who it is forA prompt engineer or art director who must ship a frame with exact text, counts, and lighting.
The problemShort prompts leave most decisions to the model, which makes precise frames hard to repeat.
The outcomeThe user has a detailed brief modeled on a craft case, adapted to their own subject and checked against the output.
What you need first
- A web browser
- Access to the Callirra gpt-image-2.5 form for generating images from a deep link
Steps
- Open the Craft cases section and pick the case closest to your target, since these run from 400 to 4,500 words.
- Read the prompt file for that case to see how canvas, light, materials, exact text, and counts are specified.
- Replace the subject and details with your own, keeping the same structure.
- Generate the frame on Callirra and compare it against the case frame to check the level of control you got.
Prompt for your agent
Using the Craft cases in callirra-ai/gpt-image-2-5-prompt-atlas as a template, write a brief for <target image> that specifies canvas, light, materials, and exact text. Output the finished brief. Verify it follows the structure of the closest case and contains no em dashes.
Caveat. The craft cases show how detailed briefs were written for this atlas, and a long brief does not guarantee a correct frame on every render.
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer