babyvideo.ai
Back to all guides
2026-01-12

How to Edit Images with AI

Image-to-image is best when you already have a base image and want controlled edits. Start with a clear source image and a single, specific instruction.

Quick Steps

A fast checklist you can follow in under a minute.

Open the tool
  1. 1
    Open the Image to Image tool.
  2. 2
    Upload a source image.
  3. 3
    Write one clear edit instruction (e.g., expand the frame, change style, fix background).
  4. 4
    Generate and compare variations.
  5. 5
    Iterate by refining the instruction (avoid conflicting requests).
Tool options you can choose
Aspect ratio:match_input_image1:116:99:163:22:34:55:43:44:3
Resolution:match_input_image0.5 MP1 MP2 MP4 MP
Format:webpjpgpng
  • Options vary by model (Nano Banana vs Flux).

Tutorial Examples (with prompts & settings)

Each example below is pre-selected for this guide (not random).

Example 1

Expand / complete an image (prompted edit)

Image-to-Imageimage-to-image
How to use this example
  1. 1.Open the tool.
  2. 2.Choose mode: Image-to-Image.
  3. 3.Upload the input shown below.
  4. 4.Use the keywords (or full prompt) and pick settings.
  5. 5.Generate and iterate (crop/lighting/prompt) if needed.
Inputs
Source
Prompt keywords
Complete the image so that the entire body appears on the screen
Show full prompt (copy)
Complete the image so that the entire body appears on the screen.
Settings (used in this example)
Open tool
Example 2

Style transform (example)

Image-to-Imageimage-to-image
How to use this example
  1. 1.Open the tool.
  2. 2.Choose mode: Image-to-Image.
  3. 3.Upload the input shown below.
  4. 4.Use the keywords (or full prompt) and pick settings.
  5. 5.Generate and iterate (crop/lighting/prompt) if needed.
Inputs
Image
Settings (used in this example)
Open tool

Tips

  • One instruction per run reduces weird artifacts.
  • If the subject changes too much, simplify the prompt and keep the edit small.
  • Use a clean source image for better results.

FAQ

How is image-to-image different from text-to-image?▼
Image-to-image starts from an existing image, so it’s more controllable for edits. Text-to-image starts from scratch and is more variable.
Can I expand the canvas (outpaint)?▼
Yes—use a prompt like “extend the image to show the full body” and keep the style consistent with the source.
Why does the style change too much?▼
Your prompt may be too broad. Use fewer style words and focus on the specific edit you want.

Ready to generate?

Open the tool and reuse the prompts/settings above.

Open the main tool
How to Edit Images with AI (Image-to-Image)