Luma AI image and video generator

Generate images, then bring them to life as video — all in one place. Luma Uni-1 creates from text and reference images with reasoning-first precision. Luma Ray 3.2 animates stills into cinematic 1080p clips with keyframe control and HDR output. Both models are on Artlist, covered by our license.

Man in dark coat stands on skyscraper rooftop at dusk, city lights stretching to the horizon — generated with Luma AI on Artlist

What is Luma AI?

Luma AI is a creative AI company that builds its own generation models, Uni-1 and Ray 3.2. Uni-1 handles image generation with multi-reference control. Ray 3.2 produces cinematic video with frame-level direction. Before generating, both models consider your prompt and produce outputs that align with your intent rather than just your keywords.

Luma Uni 1

Luma Uni-1: reasoning-first image generation

Most image models generate first and hope for the best. Uni-1 reasons first — parsing your prompt for spatial relationships, object counts, and visual logic before a single pixel renders. Add up to 9 reference images to control style, character, and composition. Use Standard for iteration, Max when the output needs to be final.

Luma Ray 3.2: cinematic AI video from text and images

Ray 3.2 generates director-grade video clips up to 20 seconds at native 1080p with HDR. Describe camera movement, pacing, and mood in plain language, or upload a still and let Ray animate it. Place up to 16 keyframes per clip for precise narrative control. Seamless 5-second loops are built in.

Luma Ray 3.2

What you can create with Luma Uni-1

Uni-1 reasons through every prompt before generating a single pixel. These are the capabilities that set it apart.

  • Reasoning-first text-to-image

    Uni-1 studies your prompt, plans composition, and then generates. That means fewer artifacts, correct anatomy, accurate object counts, and legible text rendering, including multilingual characters. Structured prompts up to 6,000 characters.

  • Multi-reference generation

    Assign roles to up to 9 reference images: style, character, composition, and lighting. Lock character identity across outputs or combine style and texture from separate sources. Build consistent visual sets without re-prompting from scratch.

  • Image editing with natural language

    Describe changes in plain English. Change backgrounds, change clothes, and adjust lighting. Uni-1 modifies the image while preserving composition and layout. Edit like a compositor, not a retoucher.

  • Standard and Max quality tiers

    Use Standard for fast iteration and exploration. Switch to Max for hero-quality finals at higher resolution. Both tiers share the same reasoning engine. Max pushes fidelity further when the output needs to be flawless.

What you can create with Luma Ray 3.2?

Ray 3.2 handles every step from prompt to polished clip. Here's what it does and what that means for your workflow.

  • Text-to-video generation

    Write a shot description with subject, action, setting, camera angle, and mood. Ray generates a clip that matches. Not keywords or tags, but actual cinematic direction in plain language. Up to 20 seconds per generation at 1080p.

  • Image-to-video with scene intelligence

    Upload any still, whether it's a photo, illustration, or 3D render. Ray reads the scene before animating: subject position, depth, and lighting direction all inform how motion is applied. Your image stays intact while everything around it comes alive.

  • Keyframe control for precise timing

    Place up to 16 keyframes in a single clip to choreograph exact narrative beats, camera paths, and visual progressions. Match storyboards and client briefs frame by frame without needing a timeline editor.

  • Seamless loops and HDR output

    Generate 5-second loopable clips for backgrounds, social assets, or motion graphics. The native HDR pipeline with 16-bit EXR output keeps lighting and materials physically accurate for post-production workflows.

Frequently asked questions

Luma Uni-1 is a reasoning-first AI image model that generates and edits images from text and reference images. It supports multi-reference control with up to 9 images, accurate text rendering, and precise spatial placement.

Luma Ray 3.2 is an AI video generation model that creates cinematic clips from text prompts or reference images. It supports 1080p HDR output, keyframe control, seamless loops, and clips up to 20 seconds.

Yes. When you generate with Luma AI on Artlist, outputs are covered by Artlist's universal license for commercial use, including ads, social content, and client projects.

Both share the same reasoning engine. Standard is optimized for speed and iteration. Max produces higher-fidelity output at roughly 2.5x the cost. Use it for final, hero-quality assets.

Yes. Generate a still with Uni-1, then use it as a reference image in Ray 3.2 to animate it into video. This image-to-video workflow keeps your subject, lighting, and composition intact.

Uni-1 supports 9 aspect ratios from 3:1 to 1:3. Ray 3.2 supports 12, including 21:9 for cinematic widescreen and 9:16 for vertical social video.

Yes. Ray 3.2 generates seamless 5-second loops, ideal for backgrounds, social assets, and motion graphics. Looping is available at the 5-second duration.

Still have questions? We're here to help.