Reve v1.5 is here. We’re releasing an early version of our latest image model today, and we’re already a top three model on Arena. Create any image you can dream of in 4K resolution.

Reve v1.5 is our most capable text-to-image model yet. It generates stunning high quality images with a significant leap forward in detail preservation, lighting, and world knowledge. Whether you're crafting images of a specific place, exploring a creative vision, or just curious about what's possible, this model can render your ideas.

What's New

4K resolution. Every image is rendered with the clarity and sharpness to hold up at full size. We unlock the full potential of high resolution generation by operating in native pixel space, as opposed to the common practice of using a "latent autoencoder" that loses precision. Create high quality output that’s print-ready and keeps the details in focus.

Better detail and lighting. We've made meaningful improvements to how the model handles fine textures, complex scenes, and the nuanced interplay of light and shadow.

Improved world knowledge. Reve v1.5 has a deeper understanding of real places, objects, styles, and concepts, which means prompts are interpreted more accurately and creatively.

What’s Coming

This is an early release, and there’s more coming soon. We’re continuing to refine our model, which will soon allow for editing images and attaching references.

We're proud of where v1.5 stands, and we can't wait for you to see what comes next.

Try Reve v1.5 at reve.com.