AI images that don't feel like your brand are almost always a DNA problem, not a prompt problem. The good news is that fixing the DNA fixes every future asset, not just this one.
Work through these in order. The first issue you find is usually the dominant one.
1. The Look Is Generic
Symptom: outputs feel like stock photography. Lighting is flat, palette is muddy, composition is centered and dull.
Likely cause: Brand DNA visual style is empty or thin. The model has no reference for what your photography actually looks like.
Fix: Settings → Brand DNA → Visual Style. Upload 6–10 reference images of your best past work. Add explicit color palette hex values. Note your typical lighting (golden hour, studio softbox, available light, etc.). Re-generate.
2. The Product Is Wrong
Symptom: the product in the image is roughly the right shape but the label is wrong, the color is off, the variant is missing, or the packaging looks made up.
Likely cause: Product DNA gap. The SKU exists but doesn't have packaging imagery, or the imagery is too low-res for the model to reproduce accurately.
Fix: Product DNA → the affected SKU → upload high-res front, side, and back imagery. Add the variant explicitly. If you have a packaging artwork file, upload it as the label reference.
For high-stakes assets, also flip to ULTRA mode. ULTRA's QA pipeline catches packaging accuracy regressions that Standard misses.
3. The Voice Is Right but the Visual Is Off
Symptom: copy on the post or caption sounds correct, but the image clashes — wrong style, wrong era, wrong audience.
Likely cause: Brand DNA captures voice well but visual style is inherited from defaults.
Fix: same as #1. Brand DNA visual style is a separate sub-section from voice, and many teams set up voice and forget visuals.
4. Hands, Faces, or Text Look Broken
Symptom: classic AI artifacts — extra fingers, melted text on packaging, faces that drift between variants in a series.
Likely cause: Standard mode shipped without QA. This is what ULTRA was built for.
Fix: re-generate the affected asset in ULTRA mode. The QA pipeline catches anatomy, text rendering, and cross-frame consistency. Keep ULTRA for the assets that matter; Standard is fine for iteration.
5. The Brief Is Vague
Symptom: every regeneration produces a wildly different output, none of which feel right.
Likely cause: the brief is asking the model to make too many decisions. "Make a social post" gives it nothing to anchor on.
Fix: tighten the brief. Specify: product, audience persona, channel, format, mood, and one or two must-includes. The brief inherits Brand and Product DNA — you don't need to re-state brand context, but you do need to set the scene.
6. Cross-Brand Bleed
Symptom: assets for Brand A include cues from Brand B (palette, mascot, voice).
Likely cause: brand context wasn't fully isolated, or the wrong brand is selected in the workspace switcher.
Fix: confirm the workspace switcher shows the intended brand. If it does and bleed continues, see Brand DNA — Advanced for multi-brand isolation rules.
Still Not Working?
Capture three example outputs and reach out to support with the brief, the DNA snapshot, and the result. The diagnostic loop is fastest with concrete cases.
