Image to Image

Guide AI generation with visual input. Instead of starting from scratch with text, use a reference image to define the structure, colors, or composition of your final output. WaveSpeed gives you precise control over the transformation process, enabling consistent style transfer and rapid iteration.

Transform Image Now View API Docs

How to Use Image to Image

The process combines visual guidance with text instructions to achieve the exact result you need.

Step 1: Upload Reference Image

Provide a source image. This acts as the blueprint for the AI, determining the basic shapes and layout.

Step 2: Add a Text Prompt

Describe the changes you want to make. For example, "convert to cyberpunk style," "change day to night," or "make the product metallic."

Step 3: Set Denoising Strength

This is crucial. A low value (e.g., 0.3) keeps the output very similar to the original image. A high value (e.g., 0.8) allows the AI to make drastic changes while only loosely following the original structure.

Step 4: Generate

Process the transformation instantly using our optimized infrastructure.

Transformation Examples

Explore common workflows where using an image prompt yields better results than text alone.

Style Transfer & Creative

Sketch to Realistic

Original Input (Before): A rough pencil sketch of a modern house design.

AI Result (After): Photorealistic Architectural Render. The lines of the sketch are transformed into realistic materials like concrete, glass, and wood under natural light.

Anime Conversion

Original Input (Before): A standard photograph of a person in a city.

AI Result (After): High-Quality Anime Illustration. The person and background are redrawn in a Japanese animation style while maintaining the original pose and setting.

Variation & Refinement

Product Variations

Original Input (Before): A photo of a sneaker in white.

AI Result (After): Colorway Generation. The same sneaker rendered realistically in red, blue, and black, keeping the lighting and shadows identical.

Composition Guide

Original Input (Before): A crude collage or basic 3D block-out showing object placement.

AI Result (After): Finished Scene. A detailed, fully rendered image that strictly adheres to the layout defined in the rough input.

Q & A

What is Image to Image (img2img)?

Image to Image is a generation method where an existing image is used as the starting point for the AI, rather than random noise. The AI takes the colors, shapes, and structure of the input image and modifies them based on your text prompt.

How is this different from Text to Image?

Text to Image creates something from nothing based only on words. Image to Image uses visual reference to provide much tighter control over the composition and structure of the final output, making it better for iterating on specific designs.

What models support Image to Image?

Most major image generation models on WaveSpeed support this mode, including FLUX.1 (dev/schnell) and Stable Diffusion XL (SDXL). We also support specialized tools like ControlNet for even more precise structural guidance.

What does "Denoising Strength" mean?

Denoising strength (sometimes called "image strength" or "creativity") controls how much the AI is allowed to deviate from your input image. Lower values preserve more of the original image; higher values allow for more creative interpretation.

Can I use Image to Image via API?

Yes. Our unified API supports image-to-image endpoints. You can send an image URL or base64 data along with your prompt to generate variations programmatically.