Reve Ships v1.5 Model, Annotation Tools, and a Redesigned Interface

Reve has shipped three updates in quick succession: a new image model, a set of annotation tools for precise editing, and a full interface redesign. Together they represent a significant step forward for the AI image generation tool that has quietly moved into the top tier of text-to-image tools.

Reve v1.5: 4K Native Resolution and Improved World Knowledge

Reve v1.5 is the company's most capable text-to-image model to date and has already reached a top-three ranking on Arena, the community-driven AI evaluation platform.

The headline feature is 4K resolution generation. Rather than using a latent autoencoder — the compression technique common across most modern image models — Reve v1.5 operates in native pixel space. This means the model retains fine detail through the generation process rather than reconstructing it from a compressed representation. The output is print-ready and holds up at full size without visible degradation.

Beyond resolution, the release includes meaningful improvements to lighting and detail handling. The model produces better results with fine textures, complex scenes, and the interplay of light and shadow. World knowledge has also been expanded, making the model more accurate when interpreting references to real places, styles, and objects. Prompts map more reliably to intended results.

Reve notes this is an early release of v1.5. Editing and image reference capabilities are in development and expected to ship soon.

Annotate Anything: Three Tools for Direct Image Control

Annotations give users a way to communicate changes visually rather than through text alone — marking directly on an image to specify what should change and where.

The feature ships with three tools:

Spotlight pins a region of an image for targeted editing. You drag a selection box to define an area, then describe what should change within it — useful for shifting the mood of a background, adjusting lighting in a specific zone, or modifying the contents of a defined space.

Draw lets you sketch a rough shape directly on the image and describe what it should become. A scribbled oval becomes a boulder. A stick figure becomes a person in a red coat. You control placement; the model handles the rendering.

Select snaps to existing objects in the scene. It's built for precise targeting — moving, resizing, replacing, or removing a specific element without disturbing the surrounding image. Select also supports collaging, letting you add objects to an existing scene.

The tools are available now for any image in Reve.

A Redesigned Interface Built Around the Canvas

The redesign consolidates controls into a single panel on the right side of the screen, replacing the previous multi-sidebar layout. Tools, settings, and controls are now in one place.

The consolidation serves a practical purpose: with fewer panels competing for space, the canvas takes up more of the screen. The expanded layout keeps the focus on the image being created rather than the interface around it.

Object editing is also more direct in the new design. Selecting any element in the image surfaces editing controls immediately, with natural-language description driving the changes. The interaction model is designed to feel less like navigating a tool and more like describing what you want.

Reve is positioning itself as a tool for designers and creative professionals who need precise control over AI-generated imagery — not just generation, but iterative editing. The combination of native-pixel 4K output, annotation-driven editing, and a streamlined interface puts it in direct competition with tools like Midjourney, Adobe Firefly, and Flux for professional use cases.

Try Reve v1.5 at reve.com.