AI Image Editor

AI Image Editor combines prompt-based image editing with common image editing utilities such as retouching, upscaling, and watermark removal. It is designed for workflows where users want to guide edits using natural language while retaining control through familiar image editing tools.

The editor follows a project-based workflow, that let users to iteratively edit, refine, and manage images.

AI Image Editor supports both global and localized edits, reference images, and model selection for generation.

What you can do:

Edit images with natural language prompts (global or localized)
Retouch specific areas using brush or smart selection
Adjust camera angles and perspectives
Resize images by expanding with AI or cropping manually
Remove or replace backgrounds with prompts, images, or colors
Upscale images up to 8x resolution
Remove watermarks automatically or manually
Adjust brightness, saturation, contrast, and warmth

Use AI Image Editor when you need both creative AI-powered edits and precise control over specific regions.

Quick Start

Create a project and upload an image
Choose your edit type:
- Generate: AI-powered edits using prompts
- Retouch: Target specific areas with brush or smart select
- Camera Angle: Adjust perspective and framing
- Resize: Expand or crop image dimensions
- Background: Remove, replace, or add shadows
- Utilities: Upscale or remove watermarks
Apply edits and preview results
Refine by adjusting or applying additional edits
Export your final image

tip

Start with simple prompt-based edits to see how the AI interprets your instructions, then use localized tools for precision.

Use Cases

AI Image Editor excels at workflows combining creative AI edits with precise control.

Ideal for:

Product photography: Remove backgrounds, adjust lighting, and retouch specific details
Portrait enhancement: Retouch specific features, adjust backgrounds, and fine-tune colors
Creative image manipulation: Add or remove elements with prompts, then refine with precision tools
Image restoration: Remove watermarks, upscale resolution, and adjust visual properties
Marketing assets: Replace backgrounds, add shadows, and ensure visual consistency
E-commerce imagery: Clean backgrounds, enhance product details, and standardize appearance

Best results when:

You combine AI edits with manual adjustments
You iterate on prompts for creative changes
You use localized tools for precision
You need both flexibility and control

Projects in AI Image Editor

Work in AI Image Editor begins by creating a project.

A project acts as a container for:

Uploaded images
Generated and edited outputs
Prompts and visual edits
Iterative changes applied over time

Projects allow users to experiment with multiple edits and refinements without losing previous context.

Generate

The Generate tab provides access to all AI-powered image editing capabilities. These features allow users to describe desired changes using text prompts and apply edits either globally or to specific regions of an image.

Prompt-Based Editing

The Prompt option lets users describe the change they want to apply to an image using natural language.

Prompts can be used to:

Add, remove, or replace elements
Modify visual attributes such as style or lighting
Perform creative or corrective edits

tip

Be specific about what you want to change. Instead of "make it better," try "brighten the sky and add warm sunset colors."

Localized Edits (Visual)

The Visual option enables localized prompt-based editing.

Using Visual edits, users can target specific regions of an image and apply prompts only to those areas. This allows for more controlled edits without affecting the entire image.

Localized prompts are useful for fine-grained changes such as editing a specific object, background area, or detail.

Reference Images

Users can add a reference image to guide the editing process.

Reference images can influence aspects such as:

Style
Composition
Color treatment

Reference images supplement the prompt and camera settings rather than replacing them.

note

Reference images influence style and composition but don't guarantee exact matches. Use them to guide the overall aesthetic.

Model Selection

Users can choose which model to use for image generation.

Available models:

Nano Banana: Consumes 3 credits per generation
Nano Banana Pro: Consumes 11 credits per generation

Model selection affects output characteristics and credit usage.

Retouch

Retouch lets users make targeted edits to specific areas or objects within an image. Unlike global edits, retouching limits the scope of changes to selected regions, providing more precise control.

Retouch supports four methods: Brush Select, Smart Select, Replace, and Erase.

Brush Select

Brush Select lets users manually define the area to be edited.

When using Brush Select:

Users can adjust the brush size to control selection precision
The desired area or object is selected by drawing directly on the image
A text prompt is provided to describe the intended change

The edit is applied only to the brushed region. This method is useful when automatic detection does not accurately capture the desired area or when fine-grained control is required.

Smart Select

Smart Select automatically detects distinct objects or regions within the image.

When using Smart Select:

The editor identifies selectable objects or areas
Users choose one or more detected regions
A text prompt is provided to describe the intended change

Smart Select is effective for quickly targeting clearly defined subjects such as people, objects, or prominent elements without manual selection.

Replace

Replace lets users change specific areas of an image using a combination of manual selection and prompt-based editing.

When using Replace:

Users draw a selection over the area they want to change
A text prompt describes the desired replacement or modification
The edit is applied only to the selected region

Replace is useful for targeted transformations where you want to specify both the exact area and the desired change, such as replacing an object, changing colors in a specific region, or modifying textures.

Erase

Erase removes the selected area from the image.

When using Erase:

Users draw a selection over the area to remove
The selected region is erased from the image
No prompt is required

Erase is useful for removing unwanted objects, cleaning up specific areas, or creating transparent regions in the image.

Retouch Behavior

Retouch edits are localized to the selected region
Each retouch produces unique results, allowing creative exploration
Re-running a retouch with the same prompt may yield different outcomes
Retouch operations consume credits according to the selected model

When to Use Retouch

Retouch is well suited for:

Editing or refining specific objects
Correcting localized visual issues
Making detailed changes without affecting the entire image
Combining manual control with prompt-based editing

note

Notes on retouch methods:

Brush Select provides maximum control but requires manual effort
Smart Select prioritizes speed and convenience when objects are clearly detectable
Replace combines manual selection with prompt-based transformation
Erase provides quick removal of unwanted areas

Resize

Resize lets users change the dimensions of an image through expansion or cropping. This feature provides control over image composition, aspect ratio, and canvas size without requiring a complete regeneration.

Two resize methods are available: Expand and Crop.

Expand

Expand increases the canvas size of an image, allowing AI to intelligently fill the new areas based on the existing image content and an optional text prompt.

Users define the expansion area by dragging the outer box around the image. The box can be adjusted freely or set to match predefined aspect ratios, providing flexibility in how the image dimensions change.

When expanding an image, users can optionally provide a text prompt describing what should appear in the newly created areas. The AI uses this prompt along with the existing image content to generate coherent extensions that blend naturally with the original image.

How Expand works:

Users start by selecting the Expand option and adjusting the outer boundary box to define how much additional canvas should be added around the image. The expansion can be applied uniformly or focused on specific sides of the image.

Predefined aspect ratios are available for quick expansion to common dimensions. Alternatively, users can manually adjust the expansion area to achieve custom dimensions.

After defining the expansion area, users can enter a prompt describing the desired content for the new regions. For example, when extending a landscape image upward, a prompt like "blue sky with white clouds" guides the AI to fill the expanded area appropriately.

Example use cases:

Expand is particularly useful when an image needs additional context or when changing aspect ratios while preserving the original subject. A portrait photo can be expanded horizontally to create a landscape composition, with the AI filling the new side areas based on the existing background. Similarly, a product photo can be expanded vertically to include more environmental context above or below the subject.

tip

Use specific prompts when expanding to guide what appears in the new areas. Descriptive prompts like "wooden floor texture" or "gradient blue sky" help the AI generate appropriate content that matches your vision.

Crop

Crop reduces the canvas size by removing portions of the image. Unlike Expand, Crop is a non-AI operation that provides immediate, deterministic results without requiring generation.

Users adjust the crop box to select the desired area of the image to keep. Everything outside the crop box is removed. The crop box can be adjusted freely or set to match predefined aspect ratios for platform-specific requirements.

How Crop works:

Users select the Crop option and adjust the crop box overlay to frame the desired portion of the image. The crop box can be resized and repositioned to achieve the exact composition needed.

Predefined aspect ratios are available to quickly crop images to common dimensions such as square (1:1), widescreen (16:9), or social media formats. This ensures the cropped image fits specific platform requirements without manual calculation.

Once the crop area is defined, applying the crop immediately removes the areas outside the selection. No AI generation occurs, making this operation instant and predictable.

Crop characteristics:

Crop does not use AI generation and therefore does not consume credits. It provides precise, manual control over composition and is useful when the desired framing already exists within the image.

The operation is immediate and deterministic, the same crop selection always produces the same result. This makes Crop ideal for quick adjustments and precise reframing.

Example use cases:

Crop is well suited for reframing a subject within the image to improve composition, changing aspect ratio to fit platform requirements, or removing unwanted edges or elements that appear at the borders of the image.

Resize Behavior

Expand and Crop serve different purposes and operate differently. Expand uses AI generation to create new image content and consumes credits according to the selected model. Crop is a non-AI operation that does not consume credits and provides immediate results.

Both methods support predefined aspect ratios for quick adjustments to common dimensions. Expand allows prompt-based control over the content generated in new areas, while Crop provides precise, manual control over which portions of the existing image to keep.

When to Use Resize

Resize is well suited for:

Changing image dimensions without starting from scratch
Extending images to include more context or environmental detail (Expand)
Reframing existing content to improve composition (Crop)
Adapting images to different aspect ratios for various platforms
Adding AI-generated content around existing subjects while maintaining the original (Expand)
Quickly removing unwanted borders or adjusting framing without AI processing (Crop)

Camera Angle

Camera Angle lets users alter the perceived camera perspective of the provided image. This feature reinterprets the image from different viewing angles without requiring a complete regeneration.

Users can adjust three camera parameters to modify how the image is viewed and composed.

Rotate Angle

Rotate angle adjusts the rotational orientation of the camera perspective. This parameter changes how the scene appears to be tilted or rotated, allowing users to reframe the composition without physically rotating the image file.

Vertical Angle

Vertical angle changes the vertical viewing perspective of the image. This adjustment simulates looking at the subject from a higher or lower vantage point, which can dramatically alter the perceived composition and emphasis within the scene.

Close-up

Close-up modifies how tightly the subject is framed within the image. Adjusting this parameter can create the effect of zooming in or out, changing the relationship between the subject and its surroundings.

These parameters influence how the image is reinterpreted during generation. The AI uses the adjusted perspective to regenerate the image, maintaining the core subject while adapting to the new viewing angle.

tip

Camera adjustments work best when combined with prompts. Try adjusting the angle, then describe the desired scene changes for more control over the final result.

Camera Angle Behavior

Camera Angle operations use AI generation to reinterpret the image from the adjusted perspective. The result maintains the subject and overall composition while adapting to the new camera parameters.

Each camera angle adjustment is treated as a new generated output and consumes credits according to the selected model.

When to Use Camera Angle

Camera Angle is well suited for:

Changing the viewing perspective without regenerating the entire image from scratch
Adjusting composition by modifying framing or orientation
Creating variations of the same subject from different angles
Experimenting with different perspectives to find the most effective composition

Upscale Image

Upscale Image lets users increase the resolution of an image while preserving visual details. This is useful when higher-resolution outputs are required for clarity, presentation, or downstream use.

Images can be upscaled by a fixed multiplier relative to the original image size.

Upscale Options

Users can choose one of the following upscale factors:

2x: Doubles the original image dimensions
4x: Quadruples the original image dimensions
8x: Increases the image size by eight times

Higher upscale factors produce larger images and may require more processing time.

tip

Start with 2x upscaling to preview results. Higher factors (4x, 8x) work best on images that are already sharp and well-composed.

Upscale Behavior

Upscaling increases image resolution, not the physical size of objects within the image
Upscaling does not modify the original image
Each upscale operation is treated as a new generated output and consumes credits

When to Use Upscale Image

Upscale Image is well suited for:

Improving image clarity for larger displays
Preparing images for high-resolution exports
Enhancing details after retouching or prompt-based edits

Remove Watermark

Remove Watermark lets users eliminate visible watermarks from images using AI-assisted detection or manual selection. This feature supports both automatic removal and fine-grained, user-controlled removal.

Two watermark removal modes are available: Auto and Manual.

Auto

Auto mode automatically detects and removes watermarks present in the image.

When Auto is selected, users can choose from the following options:

Remove text
Removes text-based watermarks.
danger
When enabled, all text present in the image will be removed.
Remove logo
Removes logo or graphic watermarks. This option is optimized for non-text elements and preserves other parts of the image.

Auto mode is useful when watermarks are clearly visible and consistently applied across the image.

Manual

Manual mode allows users to explicitly define which areas of the image should be processed for watermark removal.

When Manual is selected:

Users draw directly over the watermark areas to be removed
Brush size can be adjusted to control selection precision
The same advanced options are available:
- Remove text
- Remove logo

Manual mode is useful when automatic detection does not fully capture the watermark or when selective control is required.

Remove Watermark Behavior

Watermark removal is applied only to selected or detected regions
Auto and Manual modes both rely on AI-assisted processing
Results may vary depending on watermark complexity and placement
Each watermark removal operation is treated as a new generated output and consumes credits

When to use Remove Watermark

Remove Watermark is well suited for:

Cleaning images with visible text or logo overlays
Preparing assets for reuse or presentation
Removing branding or marks from reference images before further edits

Background

The Background tools let users remove, replace, or enhance the background of an image. These operations focus on separating the foreground subject from its surroundings and modifying the background without altering the primary subject.

Three background operations are available: Remove, Replace, and Add Shadows.

Remove

Remove eliminates the background from the provided image while preserving the main subject.

To improve background removal accuracy, users must specify the type of primary object present in the image. Available options include:

General
E-Commerce
Car
Human
Object

Selecting the correct object type helps the system optimize subject detection and edge separation, especially around complex outlines.

tip

Selecting the correct object type significantly improves edge detection. Choose "Human" for portraits, "E-Commerce" for products on clean backgrounds.

Replace

Replace lets users substitute the existing background with a new one. Background replacement can be performed using one of the following methods:

Prompt-Based Background

Users can describe the desired background using a text prompt. The prompt defines the environment, scene, or visual style of the new background.

Image-Based Background

Image-based background replacement uses a selected image as the new background.

Before applying the new background, the editor first removes the existing background from the image. After background removal is complete, users can select a new background image from two sources:

Preset images: Choose from background images available within the editor
Custom upload: Upload a custom background image

The selected image is used as the new background while preserving the foreground subject.

AI Image Editor: Image-based background replacement

Color-Based Background

Users can apply a solid color as the background. This is useful for clean, minimal, or studio-style outputs.

Add Shadows

Add Shadows generates shadows to visually ground the subject within the image.

Shadow generation requires the image to contain a full or lower-half human subject. Shadows will not be generated if a suitable human subject is not detected.

Available shadow types include:

Soft
Hard
Custom

For Custom shadows, users can control:

Intensity
Angle

Shadows are generated relative to the subject's position and the implied lighting direction.

Background Behavior

Background operations focus on separating foreground and background content
Replace and shadow generation depend on successful subject detection
Each background operation is treated as a new generated output and consumes credits

When to Use Background Tools

Background tools are well suited for:

Removing distractions from images
Placing subjects into new environments
Creating studio-style or minimal backgrounds
Adding realism through shadows after background replacement

Adjust

Adjust lets users fine-tune basic visual properties of an image. These adjustments provide deterministic control over the image's appearance and can be used independently or alongside AI-powered edits.

Adjustments are applied directly to the image and affect only the current version being edited.

Available Adjustments

The following image properties can be modified:

Brightness: Controls the overall lightness or darkness of the image.
Saturation: Adjusts the intensity of colors in the image.
Contrast: Controls the difference between light and dark areas, affecting overall visual depth.
Warmth: Shifts the color temperature of the image toward warmer or cooler tones.

Adjust Behavior

Adjustments are deterministic and produce consistent results
Multiple adjustments can be combined on the same image
Adjustments do not change the original uploaded image
Adjustments can be applied before or after AI-based edits

When to Use Adjust

Adjust is well suited for:

Correcting exposure or color balance
Fine-tuning images after retouching or background changes
Achieving visual consistency across multiple edits
Making subtle improvements without regenerating the image

Quick Reference

Feature	Options
AI editing	Prompt (global), Visual (localized)
Retouch methods	Brush Select, Smart Select, Replace, Erase
Camera Angle	Rotate angle, Vertical angle, Close-up
Resize methods	Expand (AI-powered), Crop (manual)
Background tools	Remove, Replace (prompt/image/color), Add Shadows
Upscale factors	2x, 4x, 8x
Watermark removal	Auto, Manual
Adjustments	Brightness, Saturation, Contrast, Warmth
Models	Nano Banana (3 credits), Nano Banana Pro (11 credits)
Shadow types	Soft, Hard, Custom
Object types	General, E-commerce, Car, Human, Object

Credit Consumption

Credits are consumed each time an image is generated or regenerated.

Nano Banana consumes 3 credits per generation
Nano Banana Pro consumes 11 credits per generation

Repeated generations using the same prompt or settings consume credits each time.