Skip to Content
Using StyleRefUse your style in AI toolsFLUX & diffusion models

FLUX & diffusion models

Use the flux format for FLUX and DALL·E, and the diffusion format for Stable Diffusion. Both place the subject first, then the style block.

FLUX and DALL·E

Use the flux format — natural-language style prose with inline hex colors. Put your subject on the first line, then the style block. No negative prompt is needed.

a hero image of a mountain lake at dawn

<compiled flux-format style prose with inline hex colors>

Stable Diffusion

Use the diffusion format — it includes weighted tags. Append your subject to the positive prompt and keep the weighted tags intact.

<compiled diffusion-format weighted tags>, a mountain lake at dawn

Tips

  • FLUX has no native style-reference feature, so a text spec is how you lock an aesthetic. Keep the same seed to hold composition steady while you vary the subject.
  • Don’t rewrite the diffusion tags into prose — the weights are part of the spec.
  • To compose subject + style in one call from an agent, use apply_style_to_prompt with target: "flux", "dalle", or "stable-diffusion". See Tools reference.

What to expect

Applying the same style block keeps the look consistent across generations within FLUX (or within Stable Diffusion). The same spec still renders differently between the two — each model interprets it its own way.

Next steps

Last updated on
MidjourneyPlans & limits