Reve 2.1

Reve 2.1 is an image editing model that transforms an existing picture based on a simple text description of the change you want. Instead of masking, layering, or manually retouching, you upload a reference image, describe the edit in everyday language, and the model reworks the picture to match your instruction. Developed by Reve, this version is built around three standout strengths: strong prompt adherence, layout intelligence, and accurate text rendering — the qualities that most often separate a usable edit from one you have to redo.

At its core, Reve 2.1 takes a reference image and a written prompt and returns an edited version. The prompt can be as casual as "Give him a friend" or far more detailed, and the model interprets your intent while keeping the rest of the image coherent. Because it understands layout, it can add, remove, or rearrange elements in a scene without breaking the overall composition. And because it renders text accurately, it's especially useful when your edit involves words on signs, packaging, posters, or product labels — an area where many editing models struggle to produce readable results.

Getting started is straightforward. You supply a reference image and a text prompt describing the change. Reference images can be provided in a wide range of common formats, including PNG, JPEG, WebP, AVIF, and HEIF, so you can work directly from photos, screenshots, or previously generated artwork. The model accepts images up to 4096 by 4096 pixels and as small as 128 pixels on a side, giving you flexibility whether you're editing a high-resolution hero shot or a small asset.

Reve 2.1 gives you a handful of practical creative controls. You can set the aspect ratio of the result from an unusually broad menu — from ultra-wide panoramic formats like 4:1, 3:1, and 21:9, through standard shapes like 16:9, 3:2, 4:3, and 1:1, all the way to tall vertical formats such as 9:16, 2:3, and 1:4. If you'd rather not decide, an automatic mode lets the model choose the aspect ratio it judges most appropriate for your request, which is handy when you want the edit to feel natural without fussing over dimensions.

You can also generate multiple variations at once — up to four images from a single prompt — which makes it easy to compare interpretations of the same instruction and pick the one that best fits your vision. Output can be delivered in PNG, JPEG, or WebP, so you can match the format to your workflow, whether you need lossless quality, smaller file sizes, or web-friendly assets.

The range of edits you can request is broad. Because the model follows prompts closely, you can add new subjects or objects to a scene, remove unwanted elements, change the setting or background, adjust styling, or modify specific details while leaving the rest intact. Its layout awareness means added elements tend to sit believably within the existing composition rather than looking pasted on. And its text-rendering ability makes it a strong choice for edits where legible wording matters — updating a label, changing a headline, or placing accurate text into a design.

Reve 2.1 is well suited to a wide spread of creative professionals. Designers can iterate on layouts, swap copy, and refine mockups without rebuilding a file from scratch. Marketers and social media creators can adapt a single image into multiple aspect ratios for different platforms, or spin up variations for A/B testing. Product and e-commerce teams can adjust packaging, signage, or promotional imagery while keeping text readable. Illustrators and concept artists can extend or modify their own work, adding characters or reworking scenes with a quick prompt. Photographers can make targeted changes to a shot without opening a full editing suite.

A few practical tips help you get the most from the model. Because prompt adherence is one of its strengths, clear and specific instructions tend to produce better results — describing exactly what you want changed, and how, guides the edit more reliably than vague requests. When an edit involves text, spelling out the exact wording you want gives the model the best chance of rendering it accurately. If your first result isn't quite right, generating several variations at once is an efficient way to explore different interpretations and choose the strongest option. Choosing an aspect ratio that matches your final use — rather than cropping later — helps preserve composition and detail.

There are some considerations to keep in mind. Reve 2.1 is an editing model, meaning it always works from a reference image; it's designed to transform pictures you already have rather than to conjure images from a blank canvas. Your reference image needs to fall within the supported size range, and it should be in one of the accepted formats. The model produces one to four images per request, so if you need a large set of variations you'll build them across multiple runs. As with any prompt-driven tool, results depend on how clearly you describe your intent, and complex or ambiguous edits may take a few attempts to dial in.

In short, Reve 2.1 is a focused, capable image editor that turns written instructions into polished visual changes. Its combination of faithful prompt following, smart handling of scene layout, and dependable text rendering makes it particularly valuable for design-driven and commercial work, where getting the details — and the words — right the first time saves real time. Whether you're refining a single hero image or producing platform-specific variations at speed, it gives you a fast, flexible way to reshape existing visuals with nothing more than a clear description of what you want.