Image-to-Image Models

Overview

The Image-to-Image API provides access to 55+ models that edit, transform, and enhance existing images. These models automatically activate when you upload a reference image in the Image Studio.

Image-to-image models support single or multi-image input (up to 14 images for some models) for complex edits and style transfers.

Model Categories

Multi-Image Edit Models

Advanced models that accept multiple reference images for complex transformations.

Nano Banana 2 Edit — Up to 14 images, 1K/2K/4K resolution
Nano Banana Edit — Up to 10 images with auto aspect ratio
Nano Banana Pro Edit — Up to 8 images with professional quality
Flux Kontext Dev I2I — Up to 10 images for context-aware editing
GPT-4o Edit / Image-to-Image — 5–10 images with natural language instructions
Bytedance Seedream Edit v4/v4.5 — Up to 10 images with dreamlike style transfer
Kling O1 Edit Image — Up to 10 images with 1K/2K resolution
Flux 2 Flex/Pro Edit — 8 images with next-gen quality
Vidu Q2 Reference to Image — 7 images for cinematic composition

Style Transfer Models

Apply artistic styles and effects to your images.

Midjourney v7 I2I / Style Reference / Omni Reference — Cinematic style transfer
Bytedance Seededit v3 — Natural language style changes
Flux Redux — Reimagine scenes with new compositions
Higgsfield Soul I2I — 100+ preset styles (DigitalCam, Grunge, Y2K, etc.)
Reve Image Edit — Fantasy transformations with pose preservation
Flux Kontext Effects — 13 effects (Age Progression, Cartoonify, Weather, etc.)
Image Effects — 18 presets (Cyberpunk, Felt 3D, Lofi Pixel, etc.)
Nano Banana Effects — 17 effects (3D Figurine, Famous Art, Decades, etc.)

Enhancement & Utility Tools

Practical tools for image processing and enhancement.

AI Image Upscaler — Intelligent resolution enhancement
Topaz Image Upscale — 1×/2×/4×/8× upscaling
Seedvr2 Image Upscale — 2K/4K/8K upscaling
AI Background Remover — Clean background extraction
AI Skin Enhancer — Portrait retouching
AI Color Photo — Black & white colorization
AI Object Eraser — Remove unwanted objects
AI Image Extension — Expand image boundaries

Subject & Character Models

Preserve identity while changing scenes or styles.

Flux Pulid — Identity-preserving generation
Minimax Image 01 Subject Reference — Same-subject scene transfer
Ideogram Character — Character consistency across scenes
AI Image Face Swap — Face replacement with target index selection

Specialized Tools

Domain-specific editing capabilities.

AI Dress Change — Virtual try-on
AI Product Shot — E-commerce background replacement
AI Product Photography — Professional product staging
AI Ghibli Style — Studio Ghibli anime transformation
Qwen Image Edit / Edit Plus / Edit Plus Lora — Alibaba’s multi-image editor with camera controls
Wan2.5/2.6 Image Edit — 2–3 images with ultra-high resolution (up to 5000px)
Ideogram v3 Reframe — Aspect ratio adjustment with content preservation

Popular Models

Model	Max Images	Resolution	Best For
Nano Banana 2 Edit	14	1K/2K/4K	Complex multi-image edits with Google Search
Seedream 5.0 Edit	1	Basic/High	Natural language style transfer
Flux Kontext Dev I2I	10	Flexible	Context-aware transformations
GPT-4o Edit	10	Standard	Natural language instructions
Midjourney v7 I2I	1	Variable	Artistic and cinematic edits
Bytedance Seedream Edit v4	10	1K/2K/4K	Dreamlike style with high resolution
Kling O1 Edit	10	1K/2K	Cinematic detail preservation
Higgsfield Soul I2I	1	Medium/High	100+ preset styles

Newly Added Models

These models were recently added and offer advanced multi-image capabilities.

Nano Banana 2 Edit

Family: Nano
Max Images: 14
Resolution: 1K / 2K / 4K
Aspect Ratios: Auto + 10 manual ratios

Features

Industry-leading 14-image multi-reference support
Google Search enhancement for context
Resolution scaling from 1K to 4K
Automatic aspect ratio detection
Preserves consistency across multiple edits

Example Request

{
  "model": "nano-banana-2-edit",
  "images_list": [
    "https://example.com/image1.jpg",
    "https://example.com/image2.jpg"
  ],
  "prompt": "Change her facial expression to a confident smile, and adjust the lighting to dramatic blue and purple hues.",
  "aspect_ratio": "auto",
  "resolution": "4k"
}

Seedream 5.0 Edit

Family: Seedream
Max Images: 1
Quality: Basic / High

Features

ByteDance’s latest edit model
Natural language style transfer
Quality presets for different use cases
Fast inference with high fidelity

Example Request

{
  "model": "seedream-5.0-edit",
  "image_url": "https://example.com/input.jpg",
  "prompt": "Transform into a watercolor painting style, soft pastel colors, dreamy atmosphere.",
  "quality": "high"
}

Multi-Image Input

Models with maxImages > 1 accept an array of image URLs sent in order. The order matters—the model processes images sequentially.

Multi-Image Model Support

Model	Max Images	Use Case
Nano Banana 2 Edit	14	Complex scene composition
Nano Banana Edit	10	Multi-reference consistency
Flux Kontext Dev I2I	10	Context-aware multi-image
Kling O1 Edit Image	10	Cinematic multi-shot
GPT-4o Edit	10	Natural language multi-edit
Bytedance Seedream Edit v4/v4.5	10	Dreamlike multi-reference
Nano Banana Pro Edit	8	Professional multi-image
Flux 2 Flex/Pro Edit	8	Next-gen multi-reference
Vidu Q2 Reference to Image	7	Cinematic character scenes
GPT-4o Image-to-Image	5	Multi-image reasoning
Flux 2 Klein 4b/9b Edit	4	Lightweight multi-edit
Qwen Image Edit Plus	3	Camera control with references
Flux Kontext Pro/Max I2I	2	Dual-image style blend
Wan2.5/2.6 Image Edit	2–3	Ultra-high resolution

Example: Multi-Image Request

{
  "model": "flux-kontext-dev-i2i",
  "images_list": [
    "https://example.com/ref1.jpg",
    "https://example.com/ref2.jpg",
    "https://example.com/ref3.jpg"
  ],
  "prompt": "Combine the lighting from image 1, pose from image 2, and background from image 3 into a cohesive portrait.",
  "aspect_ratio": "16:9",
  "num_images": 1
}

Common Input Parameters

Image URL / Images List

Type: string (single) or array (multi-image)
Required: Yes
Format: Publicly accessible HTTPS URL or Muapi uploaded file URL For single-image models, use image_url. For multi-image models, use images_list array.

Use the Upload File endpoint to host your images on Muapi infrastructure before sending them to models.

Prompt

Type: string
Required: Varies by model (optional for some utility tools)
Max Length: 1500–3000 characters Describe the transformation you want. Be specific about what should change and what should stay the same. Example Prompts:

Good: "Replace the barista with a humanoid robot in a sleek metallic design."

Better: "Keep the café scene unchanged but replace the barista with a humanoid robot 
with chrome finish, glowing blue eyes, and futuristic design. Preserve lighting and atmosphere."

Strength

Type: float
Range: 0.0–1.0
Default: 0.5–0.6 Controls how much the output differs from the input. Higher values = more transformation; lower values = subtle changes.

Aspect Ratio

Type: string
Options: Auto, 1:1, 16:9, 9:16, 4:3, 3:4, 2:3, 3:2, 21:9, etc. Some models support auto-detection or offer 10+ ratio options.

Model-Specific Features

Higgsfield Soul I2I — 100+ Style Presets

Select from curated style presets:Photography: DigitalCam, 90s Grain, 2000s Cam, CCTV, iPhone
Fashion: Coquette core, Y2K, Gorpcore, Indie sleaze, Grunge
Eras: 1920s/1950s/1970s/1980s Decade, 2000s Fashion
Art Styles: Ghibli, Tumblr, Avant-garde, Mixed Media
Effects: Glitch, Fisheye, 0.5 Selfie, Overexposed, Foggy Morning

{
  "model": "higgsfield-soul-image-to-image",
  "image_url": "...",
  "prompt": "Transform into cinematic editorial portrait...",
  "style": "90's Editorial",
  "aspect_ratio": "9:16",
  "strength": 0.7,
  "quality": "high"
}

Qwen Image Edit Plus Lora — Camera Controls

Precise camera adjustments without prompts:

Rotate Right-Left: -90° to +90° (positive = left, negative = right)
Move Forward: 0–10 (0 = no movement, 10 = close-up)
Vertical Angle: -1 to +1 (-1 = bird’s eye, 0 = neutral, +1 = worm’s-eye)
Wide-Angle Lens: Boolean toggle

{
  "model": "qwen-image-edit-plus-lora",
  "images_list": [...],
  "rotate_right_left": 15,
  "move_forward": 3.5,
  "vertical_angle": -0.5,
  "wide_angle_lens": true
}

Midjourney v7 — Advanced Edit Controls

Speed: relaxed | fast | turbo
Variety: 0–100
Stylization: 0–1000
Weirdness: 0–3000
Weight (Omni Reference only): 1–1000 (reference influence)

Flux Kontext Effects — 13 Effect Types

Age Progression
Background Change
Cartoonify
Color Correction
Expression Change
Face Enhancement
Hair Change
Object Removal
Professional Photo
Scene Composition
Style Transfer
Time of Day
Weather Effect

Each effect requires a prompt describing the desired change.

Utility Tools (No Prompt Required)

These tools automatically process images without text prompts.

Tool	Function	Parameters
AI Image Upscaler	Enhance resolution	None
Topaz Image Upscale	Professional upscaling	`upscale_factor`: 1, 2, 4, 8
Seedvr2 Image Upscale	Ultra-high resolution	`resolution`: 2k, 4k, 8k
AI Background Remover	Extract subject	None
AI Skin Enhancer	Portrait retouching	None
AI Color Photo	Colorize B&W images	None
AI Ghibli Style	Anime transformation	None
AI Image Extension	Expand boundaries	None
Image Passthrough	Identity function	`make_input`: true

Example Use Cases

Product Photography

{
  "model": "ai-product-shot",
  "image_url": "https://example.com/watch.jpg",
  "scene_description": "on a rock, next to the ocean, dark theme"
}

Portrait Style Transfer

{
  "model": "reve-image-edit",
  "image_url": "https://example.com/portrait.jpg",
  "prompt": "A photorealistic fantasy portrait, transforming the woman into an elegant high elf. Give her long, gracefully pointed ears. Her skin has a subtle, ethereal glow. Replace her blazer with ornate elven robes made of shimmering silver fabric."
}

Multi-Image Composition

{
  "model": "nano-banana-2-edit",
  "images_list": [
    "https://example.com/ref1.jpg",
    "https://example.com/ref2.jpg",
    "https://example.com/ref3.jpg"
  ],
  "prompt": "Combine the character from image 1, the background from image 2, and the lighting mood from image 3. Maintain consistency in style and color grading.",
  "resolution": "2k"
}

API Endpoints

All image-to-image models use the unified endpoint pattern:

POST /api/v1/{model-endpoint}

Refer to the Generate Image-to-Image guide for integration details.

Model Selection Tips

For multi-image edits: Use Nano Banana 2 Edit (14 images), Flux Kontext Dev (10 images), or Kling O1 Edit (10 images).For natural language edits: Use GPT-4o Edit or Bytedance Seededit v3.For style presets: Use Higgsfield Soul I2I (100+ styles) or Flux Kontext Effects.For upscaling: Use Topaz (professional), Seedvr2 (8K), or AI Image Upscaler (automatic).For identity preservation: Use Flux Pulid, Minimax Subject Reference, or Ideogram Character.

Best Practices

Upload First: Use the Upload File endpoint to host images before sending them to models.
Order Matters: For multi-image models, the sequence of images affects the result.
Be Specific: Describe what should change AND what should stay the same.
Test Strength: Start with 0.5 and adjust based on results.
Resolution: Higher resolutions consume more credits but provide better quality.

API Client

Models

Overview

Model Categories

Popular Models

Newly Added Models

Nano Banana 2 Edit

Seedream 5.0 Edit

Multi-Image Input

Multi-Image Model Support

Example: Multi-Image Request

Common Input Parameters

Image URL / Images List

Prompt

Strength

Aspect Ratio

Model-Specific Features

Utility Tools (No Prompt Required)

Example Use Cases

API Endpoints

Model Selection Tips

Best Practices

Build docs developers (and LLMs) love

API Client

Models

​Overview

​Model Categories

​Popular Models

​Newly Added Models

​Nano Banana 2 Edit

​Seedream 5.0 Edit

​Multi-Image Input

​Multi-Image Model Support

​Example: Multi-Image Request

​Common Input Parameters

​Image URL / Images List

​Prompt

​Strength

​Aspect Ratio

​Model-Specific Features

​Utility Tools (No Prompt Required)

​Example Use Cases

​API Endpoints

​Model Selection Tips

​Best Practices

Build docs developers (and LLMs) love

Overview

Model Categories

Popular Models

Newly Added Models

Nano Banana 2 Edit

Seedream 5.0 Edit

Multi-Image Input

Multi-Image Model Support

Example: Multi-Image Request

Common Input Parameters

Image URL / Images List

Prompt

Strength

Aspect Ratio

Model-Specific Features

Utility Tools (No Prompt Required)

Example Use Cases

API Endpoints

Model Selection Tips

Best Practices