Image-to-Image

Edit and modify existing images by uploading an image and describing the changes you want. Perfect for style transfer, adding elements, changing colors, or transforming images.

How It Works

1. Enable Generate Media - Toggle the Generate Media button in the composer

2. Attach an image - Upload or drag & drop the image you want to edit

3. Model auto-switches - System automatically switches to edit variant

4. Describe changes - Write a prompt describing what to modify

5. Configure settings - Set aspect ratio and resolution

6. Generate - Click send to create the edited image

Automatic Model Switching

When you attach an image, the system automatically:

• Detects image attachment - Recognizes you want to edit

• Switches to edit variant - Changes from text-to-image to image-to-image model

• Maintains model family - Stays in the same model family (Flux → Flux Edit)

• Reverts when removed - Switches back to text-to-image when image removed

Example:

• You select "Flux 2 Max" (text-to-image)

• You attach an image

• System auto-switches to "Flux 2 Max Edit" (image-to-image)

• You remove the image

• System switches back to "Flux 2 Max"

Supported Models

All major models support image-to-image editing:

Model	Edit Variant	Best For
Flux 2 Max	Flux 2 Max Edit	High-quality edits, style transfer
GPT Image 2	GPT Image 2 Edit	Canvas editor workflows
Nano Banana Pro	Nano Banana Pro Edit	Versatile editing, multiple images
Z-Image Turbo	Z-Image Turbo Edit	Quick edits, fast iterations
Seedream v5	Seedream v5 Edit	Creative transformations, 2K resolution

Writing Edit Prompts

13MB

IMAGE GENERATION PAPER.pdf

PDF

Open

What to Describe

Focus on the changes you want, not the entire image:

1. Specific modifications - What exactly should change?

2. Style changes - Different aesthetic or look

3. Additions - New elements to add

4. Removals - Elements to remove

5. Color adjustments - Color changes or corrections

Good Edit Prompts

✅ Style Transfer:

Make this look like a vintage film photograph with warm tones and 
film grain texture

✅ Adding Elements:

Add a beautiful sunset sky in the background with warm orange and 
pink colors

✅ Color Changes:

Change the color scheme to cool blues and purples, make it more 
moody and atmospheric

✅ Specific Modifications:

Remove the person in the background and make the scene more minimal

Bad Edit Prompts

❌ Too Vague:

Make it better

❌ Describes Entire Image:

A coffee shop with people and tables

(The image already shows this)

❌ No Clear Change:

This image

(Doesn't specify what to change)

Common Editing Tasks

Style Transfer

Transform the aesthetic of your image:

• Film look: "Make this look like 35mm film with warm tones and grain."

• Painting style: "Transform into an oil painting with brushstroke texture."

• Vintage: "Give this a vintage 1980s aesthetic with retro colors."

• Cinematic: "Make this cinematic with dramatic lighting and color gradient."

Adding Elements

Add new elements to your image:

• Objects: "Add a vintage camera on the table."

• People: "Add a person walking in the background."

• Environment: "Add rain and wet streets."

• Effects: "Add lens flares and bokeh lights."

Removing Elements

Remove unwanted elements:

• Objects: "Remove the car in the background."

• People: "Remove all people from the scene."

• Text/Logos: "Remove the text overlay."

• Distractions: "Remove the power lines and poles."

Color Adjustments

Change colors and mood:

• Color correction: "Fix the white balance, remove the blue cast."

• Color grading: "Apply warm, golden hour color grading."

• Mood changes: "Make colors more vibrant and energetic."

• Style shifts: "Change to a cool, desaturated color palette."

Composition Changes

Modify layout and framing:

• Crop/extend: "Extend the sky upward to make it taller."

• Reposition: "Move the subject to the left side."

• Background: "Replace the background with a mountain landscape."

Model-Specific Behavior

Flux 2 Max Edit

Characteristics:

• High-quality edits

• Strong style transfer

• Good at maintaining image structure

• Best for professional work

Best for:

• Style transfer

• High-quality modifications

• Professional editing

• Detailed changes

GPT Image 2 Edit

Characteristics:

• Canvas editor integration

• Composition workflows

• Layer support

• Advanced editing with up to 4 input images

• Flexible aspect ratios (8 options) and quality tiers (low / medium / high)

Best for:

• Canvas editor workflows

• Complex compositions

• Multi-image editing (up to 4 images)

• Advanced features

Nano Banana Pro Edit

Characteristics:

• Versatile editing

• Fast generation

• Supports multiple images

• Good balance

Best for:

• General editing tasks

• Quick iterations

• Multi-image workflows

• Versatile use cases

Aspect Ratio Considerations

Matching Source Image

Best practice: Use the same aspect ratio as your source image

• Prevents distortion

• Maintains composition

• Preserves important elements

• Ensures proper framing

Changing Aspect Ratio

If you change the aspect ratio:

• Image may be cropped

• Important elements might be cut off

• Composition may shift

• Use with caution

Tips for Best Results

1. Be specific about changes - "Add rain" vs. "make it better."

2. Reference the image - "Change the sky" vs. "a sky."

3. Describe style clearly - "Vintage film look" vs. "old style.e"

4. Keep changes focused - One or two major changes work best

5. Use appropriate model - Flux for quality, Turbo for speed

6. Match aspect ratios - Use the same ratio as the source image

7. Iterate if needed - Make multiple passes for complex edits

Common Workflows

Quick Style Change

1. Upload image

2. Select model (auto-switches to edit variant)

3. Prompt: "Make this look [style]."

4. Generate

5. Review and iterate if needed

Adding Elements

1. Upload base image

2. Prompt: "Add [element] to [location]."

3. Generate

4. Refine if needed

Color Correction

1. Upload image

2. Prompt: "Fix the colors, remove [color cast], make it [mood]."

3. Generate

4. Compare before/after

Troubleshooting

"Changes are too subtle."

Solutions:

• Be more specific in your prompt

• Use stronger descriptive words

• Try different models (Flux for stronger edits)

• Make the change request more explicit

"Image structure is lost."

Solutions:

• Be more specific about what to preserve

• Use models that maintain structure better (Flux, GPT Image 2)

• Avoid requesting too many changes at once

• Iterate with smaller changes

"Wrong elements changed."

Solutions:

• Be more specific about what to modify

• Reference locations ("background", "foreground")

• Use Canvas Editor for precise control

• Try different phrasing

"Model didn't switch to edit."

Solutions:

• Ensure the image is properly attached

• Check that the model has an edit variant

• Some models may not auto-switch

• Manually select edit variant if needed

---

Next: Learn about Canvas Editor for advanced image editing with layers and masks.