Images · guidance updated 22 Nov 2026
Leonardo.ai prompt guide
Omni models (FLUX.1 Kontext/Max, FLUX.2 Pro, GPT Image-1/1.5, Nano Banana/Pro, Seedream 4.0/4.5) support generation and Inline Editor editing with up to 6 reference images. Stronger contextual understanding than earlier models.
How to prompt Leonardo.ai
- Start with the most important element; early words carry more weight.
- Build prompts from subject, medium, style, composition, then colour and lighting.
- Begin with a short core phrase and add descriptors (location, action, attributes).
- Name the medium explicitly (oil painting, 35mm photo, 3D render, tilt-shift photo).
- Use commas to separate distinct concepts in longer prompts for clarity.
- Quote short text to render and pick Lucid Origin or Phoenix for typography.
- Use the negative-prompt field (where offered) for unwanted artefacts like extra limbs or blur.
- Choose a preset style in settings rather than piling style words into the prompt.
- For Omni editing: name subjects explicitly (avoid pronouns), quote text to change, state what to keep using phrases like 'keep exact camera angle, position, and framing'.
- Add up to 6 reference images to Omni models with specific instructions on how elements interact.
- Break complex edits into sequential steps; use precise verbs ('change' vs 'transform').
- Preserve likeness deliberately when editing characters; specify facial features to maintain.
The shape of a good prompt
[Medium] of [subject with key attributes] [action/pose], [setting], [composition/framing], [style], [colour palette], [lighting], text "[short words]" [position]. Omni Editing: Name subjects directly, quote text to replace, state what to preserve (e.g. 'keep exact camera angle, position, and framing'). Add reference images with interaction instructions if needed. Settings: Model [Omni type], Style preset [optional], Dimensions [aspect ratio], Negative prompt [comma list, if supported], Reference images [up to 6 for Omni].
Avoid
- Long, convoluted sentences that dilute the main subject.
- Placing the key subject at the end of the prompt.
- Inline parameters like --ar or --no.
- Pronouns ('her', 'it', 'this') in Omni edit instructions; use explicit subject names instead.
- Asking for many edits in one go on Omni models; break complex changes into steps.
- Vague text descriptions in Omni editing; always quote the exact text to change.
- Overly complex or unrealistic instructions; balance creativity with feasibility.
Example
Request a cozy coffee shop poster for autumn
Vintage screen-print poster illustration of a steaming spiced latte mug on a wooden window ledge, rainy autumn evening outside, maple leaves on wet pavement, warm amber café windows glowing, centered vertical composition with space at the top, burnt orange, mustard and deep teal palette, soft cozy lighting, visible paper grain, bold cream serif headline "AUTUMN BREWS" across the top