How to Use uni-1

uni-1 is Luma AI's unified understanding and generation model — a multimodal reasoning AI that thinks before it creates. Here is how to get started.

Create Your Account

• Visit uni1.me and click "Get Started" to sign up with Google or email
• New accounts receive free credits to explore the platform
• No credit card is required for the free tier

Navigate the Interface

• The Generator page is your main workspace — enter prompts and upload references here
• Use the top navigation to access Features, Pricing, and your Account
• Generated images appear in the preview panel on the right
• Your recent generations are saved and accessible from the history section

The quality of your output depends heavily on how you write your prompt. uni-1's reasoning engine understands nuanced descriptions, so the more specific you are, the better your results.

Prompt Structure

A strong prompt typically includes three components: the subject (what you want), the style (how it should look), and the context (where or when it exists). Combining all three gives uni-1 the clearest instructions.

• Subject: "A golden retriever wearing a space helmet"
• Style: "In the style of 1970s NASA photography, film grain"
• Context: "Standing on the surface of Mars, Earth visible in the sky"

Tips for Better Results

• Be specific about lighting: "soft golden hour light", "dramatic side lighting", "neon glow"
• Mention camera details for photorealistic styles: "85mm lens", "shallow depth of field"
• Use artistic references: "in the style of Studio Ghibli", "Wes Anderson color palette"
• Specify composition: "close-up portrait", "aerial view", "centered symmetrical"
• Include negative cues when needed: "no text", "simple background", "minimalist"

uni-1 offers flexible image generation with multiple models, aspect ratios, and quality tiers. Understanding these options helps you get exactly the output you need.

Models

• uni-1 Standard — Fast generation at up to 1K resolution, ideal for exploration and iteration
• uni-1 Pro — Higher fidelity at up to 2K resolution with enhanced detail and reasoning depth

Aspect Ratios & Resolution

• 1:1 — Square format, great for social media avatars and icons
• 16:9 — Widescreen landscape, ideal for banners and desktop wallpapers
• 9:16 — Vertical portrait, perfect for phone wallpapers and stories
• 4:3 — Classic photo ratio, versatile for most use cases
• 3:4 — Vertical classic ratio, suitable for posters and portraits

Generation Process

After submitting your prompt, uni-1 first reasons about the composition, identifies key elements, and plans the layout. This reasoning step takes a few seconds but dramatically improves output quality. The full generation typically completes in 10 to 30 seconds depending on complexity and the selected model.

One of uni-1's most powerful features is the ability to refine images through natural conversation. Instead of starting over, you can iteratively improve your results.

How It Works

• Generate your initial image with a prompt
• Review the result and identify what you want to change
• Describe the change in natural language: "Make the sky more orange" or "Remove the person on the left"
• uni-1 retains full context from the conversation and applies only the requested edits
• Continue refining across as many turns as needed — context is never lost

Editing Tips

• Make one change at a time for the most predictable results
• Be specific about which part of the image to modify
• You can change style, content, lighting, and composition across turns
• If a change goes too far, ask uni-1 to partially revert: "Make that change more subtle"

Once you are comfortable with the basics, these advanced techniques will help you unlock the full potential of uni-1.

Multi-Reference Generation

Upload up to four reference images to guide uni-1's output. The model can extract and combine elements like character design, color palette, environment, and style from your references into a coherent new image.

Style Mixing

Blend multiple artistic styles using weighted modifiers. For example, "70% watercolor, 30% digital concept art" will produce a hybrid aesthetic. Experiment with ratios to find unique combinations.

Text Rendering

uni-1 can render text directly within generated images with zero spelling errors. Specify the text in quotes within your prompt, and optionally describe the typography: font style, size, placement, and color. Multiple languages are supported including Latin, Chinese, Japanese, and Arabic.

Best Practices