Image-to-Image Workflow

Transform existing images using AI.

Concept

Image-to-image (I2I) uses an existing image as starting point, modifying it based on your prompt.

Workflow

[LoadImage] → [VAEEncode] → [KSampler] → [VAEDecode] → [SaveImage]
               ↑                           ↑
[CheckpointLoader] → [CLIPTextEncode(positive)]
                  ↓ CLIP
                  [CLIPTextEncode(negative)]

Step-by-Step

Step 1: Load Image

Node: LoadImage
Upload: Your source image

Step 2: Encode to Latent

Node: VAEEncode
Pixels: from LoadImage
VAE: from CheckpointLoader

Step 3: Enter Prompt

Node: CLIPTextEncode
Text: "describe what you want to change"
Example: "same person with longer hair, artistic style"

Step 4: Set Denoise Strength

Node: KSampler
Denoise: 0.5-0.8

Lower = more like original
Higher = more creative changes

Step 5: Decode and Save

VAEDecode → SaveImage

Key Parameter: Denoise

Denoise	Effect	Use Case
0.1-0.3	Very subtle	Color adjustment
0.4-0.6	Moderate	Style transfer
0.7-0.9	Strong	Major changes
1.0	Complete new	From scratch

Use Cases

Style Transfer

Input: Portrait photo
Prompt: "oil painting style, impressionist"
Denoise: 0.5-0.7

Age/Gender Change

Input: Face photo
Prompt: "same person, younger, smooth skin"
Denoise: 0.4-0.6

Background Change

Input: Person with busy background
Prompt: "same person, clean white background"
Denoise: 0.6-0.8

Colorization

Input: Black and white photo
Prompt: "same scene, vibrant colors, natural lighting"
Denoise: 0.6-0.8

Tips

1. Start low: 0.4-0.6 for subtle changes
2. Match resolution: Keep same aspect ratio
3. Use ControlNet: For structure preservation

Advanced: With ControlNet

[LoadImage] → [Preprocessor] → [ControlNet] → [Apply ControlNet]
                                        ↓
[Checkpoint] → [CLIPTextEncode] → [KSampler]

Next Steps

• Inpainting - Fix specific areas
• Outpainting - Extend image
• ControlNet - Structure control