-
1
Start with the intended image job: product still, campaign visual, fashion color study, pin concept, social cover, reference frame, or image-to-video source frame
-
2
Choose wan2.7-image or wan2.6-image when the brief names Wan, then compare GPT Image 2, doubao-seedream-4-5, doubao-seedream-5-0-lite, or nano-banana-pro when the output needs another still-image route
-
3
Write the prompt as an image brief with subject, composition, camera distance, aspect ratio, lighting, palette, material, brand fit, reuse context, and rejection conditions
-
4
Use X-sourced examples only to understand demand, model language, campaign visuals, color-fidelity claims, and creator workflows; do not copy third-party prompt text or product-policy claims into AIPinMaker copy
-
5
Review outputs for text accuracy, artifact density, protected-character similarity, real-person likeness risk, sensitive context, brand marks, and whether the image can become a clean source frame
-
6
If motion is useful, send only the reviewed still image to Wan 2.6 I2V, Wan 2.6 T2V, Seedance 2.0, Seedance 2.0 Fast, Kling AI, or PixVerse after the image passes review
-
7
Save the model label, prompt version, source-frame notes, comparison result, rejected-output reason, and downstream video handoff decision before publishing or reusing the image