Reve 2.1 Edit API by REVE AI

Reve 2.1 — Edit

Developer: Reve AI Model ID: reve-ai/reve-2.1/edit Release Date: July 9, 2026

Overview

Reve 2.1 Edit takes a single input image plus a natural-language instruction and returns an edited image at native 4K. Rather than repainting the whole frame, Reve treats an image as a structured set of regions — an "image as code" approach in which every element is addressable and editable. That means the model can apply precise, localized changes while leaving the rest of the picture intact, and it can iterate on an image repeatedly without quality drifting.

This edit variant is the image-conditioned entry point to Reve 2.1: you provide one image and describe how it should change. Use it whenever you already have a picture and want to modify it — as opposed to generating from scratch (reve-ai/reve-2.1/text-to-image) or combining several references (reve-ai/reve-2.1/remix).

The Reve 2.1 Model

Reve 2.1 is an update to Reve's flagship image model that sharpens the model's visual intelligence — how well it reasons about structure, hierarchy, and spatial relationships before rendering. On public leaderboards it ranks second overall while using less than a tenth of the compute of the models ranked around it, and it remains the top-rated 4K image generator.

Its foundation is a "large layout model." Images are represented as structured regions that can be reasoned about much like code, which means controllability is built into the model's visual intelligence rather than bolted on afterward. Practical consequences of this design include:

• Higher visual intelligence — the model plans in finer detail and reasons more accurately about how elements relate to one another.
• Native 4K output — full 16-megapixel images (4096 × 4096 class) with sharp edges and preserved texture, generated directly rather than through an upscaling step.
• Stronger text rendering — improved legibility and extended multilingual support, including foreign scripts.
• Precision editing — because every region is independently addressable, edits target individual elements without disturbing the rest of the scene, and repeated edits keep their quality.

Key Capabilities

• Instruction-driven editing — Describe the change in plain language and Reve applies it to the supplied image, from small local tweaks to substantial reworks.
• Element-level precision — Addressable regions let the model change one element while preserving surrounding content, composition, and style.
• Best-in-class typography — Add, remove, or replace in-image text — signs, labels, packaging copy, logos — with legible, correctly spelled results across multiple languages and scripts.
• High instruction adherence — Faithfully interprets long, detailed edit instructions (up to 2,560 characters). Instructions are automatically enhanced by the model.
• Native 4K, no upscaling — Renders true 16-megapixel results directly, holding fine detail at full size.
• Built-in post-processing — Apply background removal and a large library of stylistic effects (textures, light, and color) directly in the same request.

Input

This model takes two creative inputs:

• Image — a single input image to edit, provided as a URL or a base64-encoded string. This is the picture the model modifies.
• Text prompt — a natural-language instruction describing how to edit the image (up to 2,560 characters, automatically enhanced by the model).

Exactly one image is used. To generate an image without a reference, use reve-ai/reve-2.1/text-to-image; to combine multiple reference images, use reve-ai/reve-2.1/remix. The remaining fields below control aspect ratio, output format, and optional post-processing.

Input Parameters

Parameter	Type	Default	Range / Options	Description
prompt	string	—	≤ 2560 characters	Required. Instruction describing how to edit the image. Automatically enhanced by the model.
image	string	—	URL or base64	Required. The input image to edit.
aspect_ratio	string	auto	auto, 4:1, 3:1, 21:9, 2:1, 17:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16, 1:2, 1:3, 1:4	Desired aspect ratio. auto keeps the aspect ratio of the input image.
resolution	string	4k	4k	Output resolution. Reve 2.1 renders natively at 4K (~16 megapixels).
remove_background	boolean	false	—	Keeps only the central subject and makes the background transparent (post-processing).
effects	array	[]	0–51 items	Optional list of post-processing effects (categories: textures, light, color), applied in order. Each item selects an effect and may override its parameters.
output_format	string	png	png, jpeg, webp	Format of the output image.
enable_base64_output	boolean	false	—	Return the image as a Base64 string instead of a URL (API only).
enable_sync_mode	boolean	false	—	Wait for the result before returning, so the image is delivered directly in the response (API only).

Output

Returns the edited image, either as a URL or — when enable_base64_output is enabled — as a Base64-encoded string. By default the request is asynchronous: the response provides a request ID that can be polled until the image is ready. Enabling sync mode returns the finished image directly.

Output formats: PNG, JPEG, WebP. Resolution: native 4K (~16 megapixels).

Use Cases

• Product & e-commerce retouching — Swap backgrounds, change colorways, or clean up product shots while keeping the subject faithful.
• Localized text edits — Update signage, packaging copy, or on-image labels — including translating text into other scripts.
• Creative iteration — Refine an existing image step by step, adding, removing, or restyling individual elements without regenerating the whole frame.
• Marketing variations — Produce edited variants of a hero image for different placements and formats.
• Agentic pipelines — Programmatic, element-level editing where an agent reads an image's structure and issues precise edits without a human in the loop.

Pricing

Pricing for this model is driven by what each request actually uses. Every request produces one native-4K edited image at a fixed base price; optional post-processing (background removal and stylistic effects) adds to the total. There is no separate resolution or upscaling charge, because 4K is generated natively.

SKUs

SKU	Description	Unit Price
sku_image	Base price per generated image (Edit, native 4K, no post-processing)	$0.24
sku_remove_background_mp	Per megapixel for background removal	$0.0036
sku_effect	Per applied post-processing effect	$0.0048

Formula

cost = sku_image
     + sku_remove_background_mp × 16 × (remove_background ? 1 : 0)
     + sku_effect × number_of_effects

• Base image — every request includes one sku_image charge of $0.24.
• Background removal — when remove_background is enabled, it is billed across the full 16 megapixels of the 4K image: 0.0036 × 16 = $0.0576. When disabled, this term is $0.
• Effects — each effect in the effects list adds one sku_effect charge of $0.0048; an empty list adds nothing.

Examples

Scenario	Calculation	Cost
Base edit (no post-processing)	0.24	$0.2400
Background removal	0.24 + 0.0036 × 16	$0.2976
1 effect	0.24 + 0.0048 × 1	$0.2448
3 effects	0.24 + 0.0048 × 3	$0.2544
Background removal + 2 effects	0.24 + 0.0036 × 16 + 0.0048 × 2	$0.3072

A plain edit with no post-processing costs the base $0.24. Background removal and each stylistic effect are the only add-ons, so leaving both untouched yields the base price.