Compare Two Images Online

Compare screenshots, designs, and photos pixel by pixel without uploading either image.

Image workspace

Original versus changed

Comparison content stays here

01 · Original Drop original image here or choose a local raster image Choose image

02 · Changed Drop changed image here or choose a local raster image Choose image

Choose two local images to begin.

Purpose-built review

Use the view that answers your question

Side by side checks composition, the slider exposes before-and-after changes, overlay catches alignment drift, and pixel diff quantifies what changed. Zoom all views together and step through connected changed regions without hunting across the canvas.

Slider

Drag directly on the canvas or use arrow keys for precise review.

Overlay

Fade the changed image over the original to reveal movement.

Pixel diff

Adjust tolerance for compression noise, then inspect changed-pixel metrics.

Region review

Jump between connected changes and zoom from a fitted overview to 400%.

Private export

Download the rendered review view as a PNG created in your browser.

How to compare two images online

Are my images uploaded?

No. DiffPeek decodes, draws, analyzes, and exports your selected images in this browser tab.

What happens when image sizes differ?

Images are aligned at the top-left. Pixels outside the smaller image count as changed, and the result explains the dimension mismatch.

Why use a difference threshold?

A threshold can suppress tiny pixel variations from compression or antialiasing so meaningful changes stand out.

Try a different-size dashboard comparison

Open the generated sample images. The original is 960 × 540 and the changed image is 990 × 560, so pixels outside the smaller top-left-aligned image count as changed. Switch to Pixel diff and move the 0–255 threshold; a pixel changes when its largest RGBA-channel difference is greater than the threshold.

Supported raster inputs are PNG, JPEG, WebP, GIF, BMP, AVIF, and ICO when the browser can decode them. SVG is intentionally rejected. Each file is limited to 50 MiB and 100 megapixels; analysis scales down above a four-megapixel union while reporting the scale. Animation frames, layers, vectors, OCR, and semantic visual meaning are not compared.