Fixing extractions

The auto-extract is fast, not perfect. About one in eight brands needs at least one significant edit before the DNA looks right. The product is built around that.

When to nudge a field versus re-extract

Single field wrong (a color, one voice sample): edit the field. Approve the DNA. Done. Re-running the extract will discard your edits.

Multiple fields wrong, or the brand uses a CMS theme that obscures the real palette: re-extract from a different page on the brand’s site. Product pages and a press kit page often show truer palettes than a heavily art-directed homepage.

Brand’s site is thin, new, or mostly images: run the read anyway, then fill the gaps by hand. Every DNA field is editable, so you can set palette hex values, pick type from the system list, and add voice samples yourself before you approve. Note you do need a public URL to start a brand; the editor is for correcting and completing the read, not building a DNA from nothing.

If the read is not right

Brand DNA extraction is free and you see it before you pay for anything: the brand, the ad plan and a silent preview of your first ad all arrive without a card. So if the read does not fit your brand, nothing has been spent, and you can correct it in the editor or re-extract from a different page.

If it still is not right, email hello@brandflux.ai with the workspace name and a one-paragraph note on what is off. We reply within one business day.

Common gotchas

• Italic serifs the live site uses sparingly often get missed. Add the family by hand and mark “headline emphasis only”.
• Brand colors named differently in the design system than on the live site. Rename them to match what you call them; downstream scripts will use those names.

Next: What an ad costs covers the two prices and what is free.