> For the complete documentation index, see [llms.txt](https://docs.atlas.design/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.atlas.design/atlas-ai-studio-overview/node-index/image-nodes/image-edit-nodes.md).

# Image Edit Nodes

## When to use image edit nodes

* **Refine generated images.** Take output from [Image Generation](/atlas-ai-studio-overview/node-index/image-nodes/image-generation-nodes.md) and adjust composition, style, or content with text instructions rather than re-rolling the entire generation.
* **Isolate subjects for 3D.** Use the generative edit nodes (Text + Image → Image) to remove backgrounds, isolate single objects, or clean up cluttered references before feeding into image-to-3D.
* **Apply style transfer.** Bring an existing image into a different aesthetic via Text + Image → Image with style-reference inputs.
* **Make surgical changes.** To modify a specific aspect (color, lighting, single element) without altering the rest of the image.
* **Generate variations from a reference.** Use Text + Image List → Image with multiple references to produce assets that combine aesthetic elements from several inputs.

Atlas provides powerful image-editing nodes that can accept **text + at least one image input**:

* **Text + Image → Image Node**
* **Text + Image List → Image Nodes**

These nodes share the same core idea:

* You provide an image (or multiple images) and a text instruction, and the node generates a modified image using the chosen backend.

They differ in how much freedom the model has:

* Text+ Image **→** Image and Text + Image List **→** Image nodes allow **fully generative edits**
* It can keep the **original image untouched except for the specified modification**.

## Nodes Overview

### Text + Image **→** Image <a href="#multimodal-node" id="multimodal-node"></a>

A flexible, image-conditioned generation engine.

* Requires text + image(s)
* Merges all input images with the instruction
* Allows scene transformations, object replacement, isolation
* Produces new images guided by the provided visuals
* Different backends = different image generation models

### Text + Image List **→** Image <a href="#multimodal-node" id="multimodal-node"></a>

A versatile, multi-reference image generation engine.

* Requires a text prompt + an array of up to 4 images
* Synthesizes multiple visual inputs based on complex text instructions
* Supports object isolation, composition of multiple scenes, targeted edits and perspective shifts
* Produces a single output image by referencing specific inputs (e.g., `image_0`, `image_1`)
* Can change aspect ratio to expand the scene
* Different backends (Gemini, Reve, Flux, etc.) offer varying levels of aesthetic quality and logic handling.

<figure><img src="/files/4x4MZxxlg0HIQmlro091" alt="" width="563"><figcaption></figcaption></figure>

***

## Inputs

#### Required: <a href="#required" id="required"></a>

* **Text prompt**
* **At least one image**

Optional:

* **Multiple images**&#x20;

These inputs can be manually uploaded or generated via other nodes.

## Backends (the AI model powering the node)

You may choose from multiple backend generative models. Each backend interprets images differently:

* Some are more realistic
* Some preserve structure better
* Some are more stylized
* Some adhere more strictly to prompts

Switching backend changes the artistic and structural behavior.

<figure><img src="/files/qsjMLnMvmQEeIpFaLK9T" alt="" width="563"><figcaption></figcaption></figure>

***

## Seed Control

Changing the seed gives new variations while keeping:

* the same structure
* the same instruction
* the same conditioning images

Keeping the seed = reproducible outputs.

***

## Example Use Cases:&#x20;

### Style Transformation of a Scene <a href="#a-style-transformation-of-a-scene" id="a-style-transformation-of-a-scene"></a>

Turn a normal bookshelf into a **fantasy enchanted bookshelf**.

<figure><img src="/files/Ow0i38EVrTGAZtcuEmVR" alt="" width="563"><figcaption></figcaption></figure>

### Variation with Seed Change <a href="#b-variation-with-seed-change" id="b-variation-with-seed-change"></a>

Same prompt + same image + same backends + **different seeds** = different variations.

<figure><img src="/files/P78jzAsrSR1o1ANCxhhQ" alt="" width="563"><figcaption></figcaption></figure>

### Restyling While Preserving Layout <a href="#c-restyling-while-preserving-layout" id="c-restyling-while-preserving-layout"></a>

Transform the bookshelf into an **old, damaged one** while keeping the structure.

<figure><img src="/files/cuX4qIMAHHbGX1fT82Li" alt="" width="563"><figcaption></figcaption></figure>

### Isolate an Object with Neutral Background <a href="#d-isolate-an-object-with-neutral-background" id="d-isolate-an-object-with-neutral-background"></a>

Perfect for Image → 3D workflows.

<figure><img src="/files/elYiQp5gA2dr9VvfHwgm" alt="" width="563"><figcaption></figcaption></figure>

### Multi-Image Conditioning for Object Replacement <a href="#e-multi-image-conditioning-for-object-replacement" id="e-multi-image-conditioning-for-object-replacement"></a>

Replace specific items on a shelf using three reference object images.

<figure><img src="/files/iXmj5IcO06mzCSusJckK" alt="" width="563"><figcaption></figcaption></figure>

#### Adding Details to a Generated Asset <a href="#a-adding-details-to-a-generated-asset" id="a-adding-details-to-a-generated-asset"></a>

Example: adding decorative elements or small props without changing the underlying concept.

<figure><img src="/files/0XsUmeExGDDSWOl7BzVQ" alt="" width="563"><figcaption></figcaption></figure>

#### Removing Objects <a href="#b-removing-objects" id="b-removing-objects"></a>

<figure><img src="/files/J7Ayft5eNO891IAOHUIx" alt="" width="563"><figcaption></figcaption></figure>

#### Changing the Design Style <a href="#c-changing-the-design-style" id="c-changing-the-design-style"></a>

You can restyle the asset into a new design language (e.g., Scandinavian / futuristic / gothic) while keeping the **same layout and camera**.

<figure><img src="/files/s4F7t6EVa3FlEwpsmzLL" alt="" width="563"><figcaption></figcaption></figure>

Changing the **seed** gives controlled alternatives of the *same style*.

<figure><img src="/files/GfM5042X0huUQkUnxAKd" alt="" width="563"><figcaption></figcaption></figure>

#### Changing Image Aspect Ratio <a href="#d-changing-image-aspect-ratio" id="d-changing-image-aspect-ratio"></a>

<figure><img src="/files/L5hO5ydvMdtdHF9BNX4h" alt="" width="563"><figcaption></figcaption></figure>

## Common pitfalls

* **Using a fully-generative node when you want a precise edit.** Text + Image → Image rewrites the whole image. If you only want to change a single element, use Edit Image with Text instead. Otherwise you'll get unintended changes elsewhere.
* **Conflicting text and image guidance.** When the text instruction contradicts the reference image, output quality drops. Make sure prompts align with the visual reference, especially when using multiple reference images.
* **Over-editing iteratively.** Each pass through an edit node loses some fidelity. After 3-4 chained edits, the image often degrades visibly. For complex changes, either combine instructions into a single edit pass or re-generate from a refined prompt.
* **Forgetting Text + Image List for multi-reference cases.** When combining style, subject, and lighting from different references, Text + Image List → Image accepts multiple images. Using Text + Image → Image with a single reference and trying to describe the others in text rarely works.

## Related nodes

* [Image Generation Nodes](/atlas-ai-studio-overview/node-index/image-nodes/image-generation-nodes.md) — the upstream source of images that often feed into edit nodes.
* [2D Post Processing Nodes](/atlas-ai-studio-overview/node-index/image-nodes/2d-post-processing-nodes.md) — finishing operations (camera framing, upscaling, background removal) after edits are complete.
* [Mesh Nodes](/atlas-ai-studio-overview/node-index/mesh-nodes.md) — image-to-3D nodes consume the output of edit operations. Edit nodes are commonly used to isolate subjects before 3D conversion.
* [Utility Nodes](/atlas-ai-studio-overview/node-index/utility-nodes.md) — Describe Image and Text Generation (LLM) can refine edit instructions; Create Image List supports multi-reference Text + Image List workflows.

## Frequently asked questions

**When should I use Text + Image → Image ?**

Text + Image → Image can be used for fully generative purposes: the output is a new image guided by your prompt and reference. Use it for major transformations (style transfer, composition changes, subject swaps). Also it can be used for fully conservative purposes: the original image is preserved except where the instruction explicitly changes things. Use it for surgical edits (color shift, single element change, minor cleanup).

**Can I use multiple reference images in a single edit?**

Yes, with Text + Image List → Image. The node accepts an image array (built via Create Image List in [Utility Nodes](/atlas-ai-studio-overview/node-index/utility-nodes.md)) and uses all references when generating output. Useful for blending style references with subject references.

**Can edit nodes upscale or change resolution?**

Edit nodes preserve resolution by default. For upscaling, use the dedicated upscale nodes in [2D Post Processing](/atlas-ai-studio-overview/node-index/image-nodes/2d-post-processing-nodes.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.atlas.design/atlas-ai-studio-overview/node-index/image-nodes/image-edit-nodes.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.