How to Use Uni-1
Four steps from concept to finished image. The Uni-1 image generator handles the complexity — no design skills required.
Describe Your Vision
Type a detailed prompt into the Uni-1 image generator. Describe subjects, style, mood, and composition in natural language — the model’s multimodal reasoning handles the rest.
Be as specific as you like: mention camera angle, lighting conditions, color palette, and artistic style. Uni-1 AI decomposes your instructions and resolves spatial constraints before rendering a single pixel.
Upload References (Optional)
Add up to 8 reference images for identity preservation, style matching, or pose guidance. Uni-1 grounds your output in the source material, not random interpolation.
Reference-based generation is where Luma Uni-1 truly shines — ranked #1 in this category. Upload character sheets, mood boards, or product photos and watch the model maintain consistency across every output.
Review and Refine
Uni-1 AI generates results in seconds. Use multi-turn refinement to adjust lighting, swap backgrounds, or tweak specific regions through follow-up prompts.
Unlike traditional AI image models, Uni-1 supports native multi-turn editing. Simply describe what you want changed — no need to regenerate from scratch. This iterative workflow saves both time and credits.
Download Your Image
Export at up to 2048px resolution. All files are auto-deleted within 1 hour — your creative assets stay private.
High-resolution output starts at approximately $0.09 per image. Standard resolution is available on the free tier. All generated files are yours to use commercially.
Before You Start
- Choose between Text-to-Image and Image-to-Image mode based on whether you have source material or are starting from scratch.
- For reference-guided generation, use images with a clear subject and good lighting — the model preserves what it can see clearly.
- Think about your desired output before writing the prompt: resolution, aspect ratio, and artistic direction all affect how Uni-1 interprets your instructions.
Writing Better Prompts
Lead with the subject
Start your prompt with the main subject and its most important attributes. For example, "A confident woman in a red leather jacket" gives the model a clear anchor before you layer on scene details.
Specify style and mood
Add explicit style cues like "cinematic lighting," "manga panel," "editorial photography," or "matte painting." Uni-1 supports 76+ visual styles and responds best to concrete direction.
Describe what to keep, not just what to change
When using reference images, mention which elements should stay consistent — face identity, clothing details, brand colors — so the model knows what to preserve during generation.
Prompt Examples
Use these as starting points, then adapt them to your own creative vision.
“A Japanese ramen shop at night, warm lantern glow reflecting on wet pavement, steam rising from a bowl visible through the window, cinematic 35mm film grain, moody and atmospheric.”
“Place this character in a sunlit Mediterranean courtyard. Keep the face, hairstyle, and outfit identical to the reference. Add soft golden-hour lighting and terracotta architecture in the background.”
“Change the background to a modern glass office. Keep everything else the same — same pose, same expression, same lighting direction. Make the reflection in the glass subtle and realistic.”
Pro Tip: When using reference images with Uni-1, include at least one full-body and one portrait shot. This gives the model enough identity context to maintain consistency across wildly different compositions and camera angles.
Your data is never stored.