Luma Uni-1
Multimodal reasoning. Mask-free instruction editing. 9 reference images.
Luma Uni-1 is a multimodal reasoning model: a decoder-only transformer that reasons across text and image tokens in a single stream, so it understands a brief the way it understands a sentence. On Stensyl it powers both generate and edit. Create stills at 10 credits per image, step up to the Max quality tier for hero-grade output, or rework any image with a plain written instruction and no mask via Luma Uni-1 Edit. Up to 9 reference images, optional web-grounded generation, strong typography, and style control that runs from photoreal to manga. Available on every plan, including Free.
Example outputs
“A poster for a design week exhibition titled 'MATERIAL FUTURES', bold grotesque typography, corten steel texture background, Swiss grid layout”
“A manga panel: a young architect on a rooftop at dusk overlooking a sprawling city, dramatic perspective, screentone shading, speech bubble reading 'It starts here.'”
“A photoreal product still of a matte ceramic pour-over coffee set on travertine, soft directional window light, editorial composition”
“A character sheet from 9 references: the same woman in studio portrait, full-length editorial, and candid street photography, consistent identity and wardrobe”
“A web-grounded image of this season's trending interior palette applied to a Scandinavian living room, magazine photography style”
“An edit instruction: change the storefront signage to read 'ATELIER NORD' and shift the awning from red to sage green, keep everything else identical”
How it works
Describe your vision
Type a detailed prompt or upload a reference sketch, photo, or mood board.
Choose your settings
Pick your resolution and aspect ratio. See the credit cost before you generate.
Generate in seconds
Your image is delivered in seconds. Download, iterate, or pipe into video.
Ready to create with Luma Uni-1?
Jump into the Studio and start generating. Plans from £10/month.
One model that reasons, generates, and edits.
Luma Uni-1 is built differently from most image models on the roster. It is a decoder-only transformer that reasons across text and image tokens in a single stream, the same way a language model reasons through a sentence. Instead of translating your prompt into a bag of visual cues, it works through the logic of the brief: what is in the scene, how the elements relate, what the typography needs to say, which style rules apply. The outcome is fewer wasted generations and images that match what you actually asked for.
The same reasoning carries into editing, and this is where Uni-1 changes the workflow. Luma Uni-1 Edit takes a plain written instruction with no mask: 'swap the timber cladding for corten steel', 'change the headline to the spring tagline', 'move the product to the left third'. The model identifies what you mean, makes the change, and leaves the rest of the image alone. On Stensyl, Uni-1 powers both the generate and edit surfaces, so you can create an image and refine it with instructions in the same model family without switching tools.
Reference handling goes deep: up to 9 reference images per generation, enough to pin down a character, a product, a location, and a style direction in one pass. Optional web-grounded generation lets the model research in real time before rendering, useful for trend-aware content, current products, and real places. Typography is a genuine strength, and style control spans the full range from photoreal to illustration to manga.
Pricing is straightforward. Standard generations are 10 credits per image, the Max quality tier is 23 credits for hero-grade stills, and Luma Uni-1 Edit is 13 credits per edit. Uni-1 is available on every plan, including Free, so the full reasoning, editing, and reference stack is open to every subscriber from day one.
Edit with instructions, not masks
Luma Uni-1 Edit removes the most tedious step in AI image editing: painting masks. Describe the change in plain language and the model locates the target, applies the edit, and preserves everything else. Material swaps, copy changes, object moves, lighting shifts. Each edit is 13 credits and runs on the same reasoning engine as generation.
Up to 9 reference images
Pin down identity, product design, location, and style in a single generation. Nine reference slots is enough to brief the model like you would brief a photographer: this person, this product, this place, this mood. The model treats references as context to reason over, not textures to copy.
Web-grounded generation
Switch on web grounding and Uni-1 researches before it renders, pulling real-time context into the image. Useful for visuals that reference current products, live trends, seasonal moments, or real locations where accuracy matters more than invention.
Typography and style range
Uni-1 renders legible, well-set type for signage, packaging, posters, and UI mockups. Style control is explicit and broad: photoreal, editorial, painterly, flat illustration, and manga among them. For design professionals, this means one model covers the brand campaign and the comic panel.
Questions about Luma Uni-1.
Why Stensyl?.
Ready to create with Luma Uni-1?
Professional image generation. Plans from £10/month.