HeadsUpAI

Reve 2.0 Introduces Layout Based Generation for Precise 4K Image Control

Reve has launched Reve 2.0, a 4K image model built on a layout-based architecture. Instead of interpreting text prompts, this system uses a Large Layout Model (a model trained for spatial reasoning) derived from Qwen to represent images as structured code. Every element is addressable, allowing for precise control over size and location.
Model
Reve 2.0
Resolution
4K
Architecture
Large Layout Model
Base Model
Qwen
Leaderboard Rank
Top 2 (Arena Text-to-Image)

This update addresses the ambiguity of text-to-image prompting, where minor wording changes cause unpredictable shifts. By adopting a layout-first approach, Reve positions generation as program synthesis—a shift similar to how ChatGPT Images 2.0 moved toward functional design. The model currently ranks top two on the Arena Text-to-Image leaderboard.

Users can generate images using instructions, layouts, or reference photos. The addressable output enables targeted edits, such as moving an object without altering the background, mirroring the consistency of Aleph 2.0. Reve 2.0 is available now via the web interface, with an API launching in about a week.

View the full update on blog.reve.com
Reve
Reve
@reve
X

Today, we’re launching Reve 2.0, the best 4K image model in the world. We invented a new way to generate and edit any image using precise layouts. For the first time, it’s possible to create images you can touch. https://t.co/mdj2xDEqfp

84retweets464likes
View on X

Still wondering? A few quick answers below.

Reve 2.0 is a 4K image generation and editing model developed by the Reve research lab. It introduces a layout-based architecture that treats images as structured code rather than just pixel grids. This allows users to have precise, addressable control over every element within a visual, from object placement to specific attributes.

Layout-based generation uses a structured description where every element has a defined location, size, and local description. Unlike text prompts that can be ambiguous, these layouts act as a backbone for the image. The model uses spatial reasoning to render pixels based on this layout, ensuring that specific instructions are followed accurately.

The Reve 2.0 API is scheduled to launch in approximately one week following the initial product release. This will allow developers to integrate the model's layout-based generation and editing capabilities into their own applications. Currently, the model is accessible through the official Reve web interface for immediate testing and image creation.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update