fal-image-to-image

What this skill does

## Quick Reference

| Task | Endpoint | Key Parameter |
|------|----------|---------------|
| Transform | `fal-ai/flux/dev/image-to-image` | `strength: 0-1` |
| Inpaint | `fal-ai/flux/dev/inpainting` | `mask_url` |
| ControlNet | `fal-ai/flux-controlnet-union` | `control_type` |
| Upscale | `fal-ai/clarity-upscaler` | `scale_factor: 2-4` |
| Remove BG | `fal-ai/birefnet` | - |
| Face fix | `fal-ai/codeformer` | `fidelity: 0-1` |

| Strength Value | Effect |
|----------------|--------|
| 0.2-0.3 | Subtle style hint |
| 0.5-0.7 | Moderate transform |
| 0.8-1.0 | Full regeneration |

| ControlNet Type | Use Case |
|-----------------|----------|
| `canny` | Edge guidance |
| `depth` | 3D structure |
| `pose` | Human pose |
| `tile` | Detail enhancement |

## When to Use This Skill

Use for **image editing and transformation**:
- Style transfer while preserving structure
- Inpainting to edit specific regions
- ControlNet for guided generation
- Upscaling low-resolution images
- Removing backgrounds
- Restoring faces in old photos

**Related skills:**
- For text-to-image: see `fal-text-to-image`
- For image-to-video: see `fal-image-to-video`
- For model selection: see `fal-model-guide`

---

# fal.ai Image-to-Image Models

Complete reference for all image transformation, editing, and enhancement models on fal.ai.

## FLUX Image-to-Image

### Basic Image-to-Image
**Endpoint:** `fal-ai/flux/dev/image-to-image`
**Best For:** Style transfer, image transformation

Transform existing images while preserving structure.

```typescript
import { fal } from "@fal-ai/client";

const result = await fal.subscribe("fal-ai/flux/dev/image-to-image", {
  input: {
    prompt: "Transform into a watercolor painting style",
    image_url: "https://example.com/photo.jpg",
    strength: 0.75,  // 0-1, how much to transform
    num_inference_steps: 28,
    guidance_scale: 3.5,
    seed: 42
  }
});

console.log(result.images[0].url);
```

```python
import fal_client

result = fal_client.subscribe(
    "fal-ai/flux/dev/image-to-image",
    arguments={
        "prompt": "Transform into a watercolor painting style",
        "image_url": "https://example.com/photo.jpg",
        "strength": 0.75,
        "num_inference_steps": 28,
        "guidance_scale": 3.5
    }
)
print(result["images"][0]["url"])
```

**Strength Parameter:**
- `0.0`: No change (original image)
- `0.3-0.5`: Subtle changes, style hints
- `0.5-0.7`: Moderate transformation
- `0.7-0.9`: Strong transformation, some original preserved
- `1.0`: Complete regeneration (ignores original)

### FLUX Pro Image-to-Image
**Endpoint:** `fal-ai/flux-pro/v1/image-to-image`
**Best For:** Production quality transformations

```typescript
const result = await fal.subscribe("fal-ai/flux-pro/v1/image-to-image", {
  input: {
    prompt: "Professional headshot, studio lighting",
    image_url: "https://example.com/casual-photo.jpg",
    strength: 0.6,
    guidance_scale: 3.5
  }
});
```

## ControlNet Models

### FLUX ControlNet
**Endpoint:** `fal-ai/flux-controlnet`
**Best For:** Precise structural control

```typescript
const result = await fal.subscribe("fal-ai/flux-controlnet", {
  input: {
    prompt: "A modern house with large windows",
    control_image_url: "https://example.com/architecture-sketch.png",
    controlnet_conditioning_scale: 0.8,
    num_inference_steps: 28,
    guidance_scale: 3.5
  }
});
```

### FLUX ControlNet Union
**Endpoint:** `fal-ai/flux-controlnet-union`
**Best For:** Multiple control types in one model

Supports: canny, depth, pose, tile, blur, gray, low_quality

```typescript
const result = await fal.subscribe("fal-ai/flux-controlnet-union", {
  input: {
    prompt: "A beautiful woman in a red dress",
    control_image_url: "https://example.com/pose-reference.png",
    control_type: "pose",
    controlnet_conditioning_scale: 0.7,
    num_inference_steps: 28
  }
});
```

**Control Types:**
- `canny`: Edge detection guidance
- `depth`: Depth map guidance
- `pose`: Human pose guidance
- `tile`: Detail enhancement
- `blur`: Blur-based control
- `gray`: Grayscale structure
- `low_quality`: Quality enhancement

### SDXL ControlNet
**Endpoint:** `fal-ai/fast-sdxl/controlnet`
**Best For:** SDXL with structural control

```typescript
const result = await fal.subscribe("fal-ai/fast-sdxl/controlnet", {
  input: {
    prompt: "A futuristic cityscape",
    negative_prompt: "blurry, low quality",
    control_image_url: "https://example.com/depth-map.png",
    controlnet_type: "depth",
    controlnet_conditioning_scale: 0.8
  }
});
```

### Canny Edge Detection
**Endpoint:** `fal-ai/flux-controlnet-canny`
**Best For:** Edge-based generation

```typescript
const result = await fal.subscribe("fal-ai/flux-controlnet-canny", {
  input: {
    prompt: "A detailed architectural drawing",
    control_image_url: "https://example.com/building-photo.jpg",
    controlnet_conditioning_scale: 0.9
  }
});
```

### Depth-Based Control
**Endpoint:** `fal-ai/flux-controlnet-depth`
**Best For:** 3D-aware generation

```typescript
const result = await fal.subscribe("fal-ai/flux-controlnet-depth", {
  input: {
    prompt: "A mystical forest scene",
    control_image_url: "https://example.com/depth-map.png",
    controlnet_conditioning_scale: 0.7
  }
});
```

## Inpainting Models

### FLUX Inpainting
**Endpoint:** `fal-ai/flux/dev/inpainting`
**Best For:** Editing specific regions

```typescript
const result = await fal.subscribe("fal-ai/flux/dev/inpainting", {
  input: {
    prompt: "A golden retriever",
    image_url: "https://example.com/photo-with-pet.jpg",
    mask_url: "https://example.com/mask.png",  // White = edit area
    num_inference_steps: 28,
    guidance_scale: 3.5
  }
});
```

```python
result = fal_client.subscribe(
    "fal-ai/flux/dev/inpainting",
    arguments={
        "prompt": "A golden retriever",
        "image_url": "https://example.com/photo.jpg",
        "mask_url": "https://example.com/mask.png"
    }
)
```

**Mask Format:**
- White pixels (255, 255, 255): Areas to edit/regenerate
- Black pixels (0, 0, 0): Areas to preserve

### FLUX Pro Fill
**Endpoint:** `fal-ai/flux-pro/v1/fill`
**Best For:** High-quality inpainting and outpainting

```typescript
const result = await fal.subscribe("fal-ai/flux-pro/v1/fill", {
  input: {
    prompt: "Seamless continuation of the landscape",
    image_url: "https://example.com/partial-image.jpg",
    mask_url: "https://example.com/outpaint-mask.png"
  }
});
```

### SDXL Inpainting
**Endpoint:** `fal-ai/fast-sdxl/inpainting`
**Best For:** SDXL-based region editing

```typescript
const result = await fal.subscribe("fal-ai/fast-sdxl/inpainting", {
  input: {
    prompt: "A beautiful flower arrangement",
    negative_prompt: "ugly, distorted",
    image_url: "https://example.com/vase.jpg",
    mask_url: "https://example.com/vase-mask.png"
  }
});
```

## Upscaling Models

### ESRGAN 4x
**Endpoint:** `fal-ai/esrgan`
**Best For:** Classic 4x upscaling

```typescript
const result = await fal.subscribe("fal-ai/esrgan", {
  input: {
    image_url: "https://example.com/low-res-image.jpg",
    scale: 4  // 2 or 4
  }
});

// Result is 4x the original resolution
console.log(result.image.url);
```

```python
result = fal_client.subscribe(
    "fal-ai/esrgan",
    arguments={
        "image_url": "https://example.com/low-res.jpg",
        "scale": 4
    }
)
print(result["image"]["url"])
```

### Clarity Upscaler
**Endpoint:** `fal-ai/clarity-upscaler`
**Best For:** AI-enhanced upscaling with detail generation

```typescript
const result = await fal.subscribe("fal-ai/clarity-upscaler", {
  input: {
    image_url: "https://example.com/image.jpg",
    scale_factor: 2,  // 1-4
    prompt: "high quality, detailed, sharp",
    creativity: 0.3,  // 0-1, how much to add
    resemblance: 0.8  // 0-1, similarity to original
  }
});
```

**Parameters:**
- `scale_factor`: 1-4, upscale multiplier
- `creativity`: 0-1, how much AI can "imagine" details
- `resemblance`: 0-1, how closely to match original
-