What’s new
- Multi-reference fusion: combines up to 10 reference images into one new output (the image-editing endpoint Nomi reaches through kie.ai caps at 8 — see Known limits).
- Editing output up to 4MP: edits keep their detail and coherence instead of degrading with each pass.
- Noticeably stronger text and layout: small text stays legible in dense layouts, infographics and UI mockups.
- Better world knowledge and prompt adherence: it follows more complex descriptions, composes more coherently, and holds brand guidelines (colors, fonts, logos) more consistently.
What it’s good for
- Commercial visuals that need to stay on-brand: consistent colors, fonts and logos.
- Text-heavy, precisely laid-out work like infographics and UI mockups.
- Multi-image composition: combining several source images into one new scene while keeping each subject recognizable.
Prompting tips
- Write subject, then action, then style, then context: put the most important details first — word order visibly affects what the model prioritizes.
- No negative prompts: the official guide states "FLUX.2 does not support negative prompts" — describe what you want, not what to avoid.
- Be specific about photography: naming a camera, lens and film stock ("shot on a Hasselblad X2D, 80mm lens, f/2.8") beats a generic "professional photo".
- Use hex codes for color: tying a hex value to a specific object ("the sofa in hex #1B6B6F") is far more precise than a color name.
- Quote in-image text and describe placement/font: wrap the text you want rendered in quotes and say where it sits and what typeface it should look like.
- Use structured JSON prompts for complex scenes: declaring scene, subjects, positions, lighting and camera separately helps with production work that needs to repeat reliably.
A minimalist product shot for a skincare brand, subject centered on a matte pedestal in hex #EDE6DB.
Shot on a Hasselblad X2D, 80mm lens, f/2.8, soft studio light from the upper left.
A small card beside the product reads "NEW · Vitamin C Serum" in elegant serif type.
Square 1:1, clean negative space on the right third for a text overlay.
Known limits
- No negative prompt support — only positive descriptions work.
- Text-to-image and image-to-image are separate model IDs: the edit endpoint Nomi reaches via kie.ai tops out at 8 reference images and supports one extra "auto" aspect-ratio option that text-to-image doesn't — the two parameter sets aren't interchangeable.
- Maximum output is 4MP, and dimensions must be multiples of 16 or the request is rejected.
Using it in Nomi
On the generation canvas, add a image card, pick “FLUX.2 Pro” in the model picker and choose a mode (Text to image, Image editing). Add your references, write the prompt and generate. In the storyboard you can pick it per shot. The first time, connect any one of the providers above under Settings → Models.
