--- title: "Introducing Layers" description: "Powered by Uni-1, Layers gives Luma customers precision editing and control over every element of any image. It's AI generates images as independent, editable layers, and it decomposes any uploaded image into layers too." canonical: "https://lumalabs.ai/news/introducing-layers" source: "https://lumalabs.ai/news/introducing-layers.md" --- # Introducing Layers _July 29, 2026_ Change one element. Preserve everything else. ![Luma layers powered by Uni-1](https://cdn.sanity.io/images/2ylxvaa2/production/90d53f4eadd8c958ee35ef9d0c97b7ff6e8e98e7-1058x592.png) ### Tired of wasted regenerations? So were we. Agency teams. In-house brand marketers. Post-production shops. Different jobs, same story every time. There's a great asset everyone loves: the exact composition, mood, shot. What's left is a quick fix. In theory. Swap the product. Change the headline. Make the background a shade less warm. Localize the copy for thirty additional markets. Until now, that fix meant regenerating the entire image and hoping those minor updates would stick, or rebuilding each image by hand in another tool. But here is the problem: full AI regenerations are fickle. Every pass through a model further degrades the original image, a photocopy of a photocopy, and the changes everyone’s approved may not survive the trip. One change shouldn't cost you everything. Luma's own in-house creative team endured this very experience for years. They'd export every image out of Luma just to make final edits, then bring it back. That's why we built Layers. Powered by Uni-1, Layers gives Luma customers precision editing and control over every element of any image. It's AI generates images as independent, editable layers, and it decomposes any uploaded image into layers too. ### How Do Layers Work? From any image, every object, every text element, every background becomes its own asset. Adjust the one you want. Leave the rest untouched. Recompose, and no one can tell where the edit was made. Swap a product. Restyle a background. Regenerate a single object. Everything else in the frame stays exactly where you left it, without leaving the Luma canvas. Change the copy, change the font color, drop in a custom font, and nothing else in the frame moves. A copy change takes two clicks, not a regeneration and a re-approval. And because untouched layers never go back through the model, every edit is a first edit: the layout you approved on day one is pixel-identical to version ten. AI video editing made simple: Decompose. Fix. Recompose. ### Native to Generation Other tools take a flat image and try to reverse-engineer it into something editable after the fact. Uni-1 doesn't work backward. It generates images as independently addressable layers from the very first pixel. Layering is native to how the model understands a scene, not a pass applied to a finished picture. ### The Same Intelligence, Applied to Uploads That same understanding runs in reverse. Upload a poster, a screenshot, a brand asset from three years ago, and Uni-1 reads it the same way it generates one: parsing what a region is and how it relates to the rest of the scene, not just clustering pixels that look alike. That distinction is what makes an edit to one layer hold up once everything recomposes. The Uni-1 tech report already showed this: training a model to generate also sharpens how it understands regions, objects, and layout. ### More Control Across Use Cases - Edit as live text: Change copy, style text, or change font or colors. Text is not baked into pixels but adjustable, down to custom brand fonts. - Rearrange per platform: Make Z-order adjustments to group, hide, or delete. Position elements precisely inside the layer editor and change aspect ratio per platform. - Extract any element: Pull any layer onto a transparent background as its own standalone asset. The product shot, the logo, the character: each one reusable anywhere, not locked inside the composite it was born in. One image becomes a whole asset library. - Mix and match across images: Stored layers become reusable ingredients that can be recombined across compositions. Drag a layer from one image into another and never start from a blank canvas again. - Recompose existing creatives: Upload a poster from three years ago, or a brand asset nobody kept the source files for, and it decomposes the same way. - Localize into every market at once: Layers scales past a single fix. Generate batch variations, like localizing ad copy across dozens of assets, while layout and style hold constant across every one. - Finish: Upscale individual layers to 2K on demand for delivery quality. Layers works on any image, Luma generations, or uploads. And the whole loop runs conversationally in Luma Agents: ask the agent to extract the layers you want, then keep chatting to change or modify any layer in the editor. Extraction, editing, and recomposition, all by conversation, in one place. ### New Workflows for Creatives The brief was: make it possible for AI to deliver an image fix as easy as the client thinks it should be. Layers now gives you the surgical edit and the batch job, in the same tool, on the same image. Start in Luma. Finish in Luma. [Layers is now live in Luma Agents. Try it today.](https://lumalabs.ai/app)