How to Keep Exact Brand Colours in AI Fashion Images (Hex to Prompt)

Quick answer: most image models don't read hex codes like #1B2A4A reliably. They read words. Describe the colour by name, undertone, depth and finish ("deep navy with a cool blue undertone, matte cotton"). Our free Palette to Prompt tool turns up to 5 hex codes and a mood into that wording.

Why hex codes fail

Models learned colour from image captions, and captions use words. A hex code is often treated as random text, so you get "a blue", not your blue.

The 4 parts of a colour description

1. Name: navy, camel, sage, cream, burgundy
2. Undertone: warm, cool, yellow-leaning, grey-leaning
3. Depth: pale, mid, deep
4. Finish: matte, satin sheen, glossy, brushed

camel wool coat, warm golden-brown undertone, mid depth, soft matte finish

Workflow for catalogue consistency

• Write the colour sentence once and reuse it in every prompt
• Use a reference photo of the real product whenever possible
• Shoot in even light (see studio lighting prompts)
• Check final images against the real sample, and correct small shifts in an editor

Colour from AI is close, not guaranteed. For product pages, always compare to the physical item.

Related

• AI photo editor: 7 edits
• Creative Workspace brand looks

Turn your hex codes into prompt words

FAQ

Can Midjourney or Gemini use hex codes?

Sometimes, but not reliably. Describing the colour in words with undertone and finish works better.

How do I keep the same colour across a campaign?

Reuse the exact same colour sentence, use a product reference photo, and keep lighting the same.

Is AI colour accurate enough for product pages?

Close, but check against the real product and correct where needed.

Next step: learn the whole workflow in a recorded workshop (see all workshops) or save with a bundle.