Model Details
Uni-1 Max generates a single, highest-fidelity image from a text prompt, with richer detail and stronger prompt adherence than the standard tier — built for hero-quality stills where one polished result matters more than a batch of variations.
## Features - Single high-fidelity output tuned for detail and prompt adherence - Optional manga styling for line-art and comic aesthetics - Web-grounded generation that can consult the web for references when enabled - Reference-guided generation from up to nine images - Flexible framing via aspect_ratio, with PNG or JPEG output
## What it's good at - Polished single "hero" images for covers, posters, and key art - Faithful rendering of detailed, specific prompts - Stylized illustration when manga mode is selected
## Style The `style` field selects the rendering aesthetic: - `auto` (default) — picks the most fitting photorealistic or illustrative style for your prompt. - `manga` — black-and-white manga / comic line-art. Choose this for paneled illustration, inked characters, or graphic-novel looks.
## Web-grounded references Set `enable_web_search` to let the model consult the web for visual references while generating — useful when your prompt names real, recent, or niche subjects it should depict accurately. Leave it off for faster, self-contained generation.
## Reference-guided generation Pass up to nine image URLs in `reference_image_urls` to steer composition, subject, or style toward your own references; the model blends their guidance with your prompt.
## Tips - Write specific prompts — Max rewards detail with fidelity. - Use manga only when you want line-art; auto covers most photoreal needs.
## Limitations - Produces one image per request (no batch variations).
To run via the ModelRunner JavaScript client: ```js import { modelrunner } from "@modelrunner/client";
const result = await modelrunner.subscribe("luma/uni-1-max", { input: { prompt: "a lighthouse on a cliff at golden hour, dramatic sky", aspect_ratio: "16:9", style: "auto", }, }); ```