Clean tools
← PNG to SVG

JPG to SVG

Convert a JPG into a scalable SVG vector directly in your browser. Drop a photo to start. No upload required.

Drag and drop an image, paste one,
or use the button below.

Tracing options

Mode

Palette: 16 colors

Detail

Fewer colors and the Smooth setting give a smaller, cleaner SVG. Logos, icons, and flat artwork trace best.

JPG to SVG

This page is set up for JPG and JPEG sources with a 16-color palette and balanced detail, a good starting point for the soft edges and slight compression noise that JPG photos carry. Drop or paste a JPG and it is traced into flat color regions, each fitted with smooth curves and written out as an SVG path you can scale or edit.

Tracing runs in a Web Worker on your own device using the Canvas API and a self-hosted tracer. Very large images are traced at a reduced working resolution, since a vector does not need every pixel, and the SVG keeps the original dimensions. Your image is never uploaded, and the output is plain SVG you can download or copy as code. No signup, no tracking.

Frequently asked questions

Why does a JPG trace less cleanly than a PNG?

JPG compression adds faint blocky noise around edges, which the tracer can pick up as tiny extra shapes. Lowering the color count or choosing Smooth merges those specks away. If you have the original PNG or a higher quality export, that will trace more cleanly.

Can I change the settings?

Yes. The preset only picks the starting mode, palette size, and detail. Change any option and the image is retraced in place, with the side-by-side preview updating each time.

More ways to vectorize

View all tools

Use Clean.tools from your AI agent

Clean.tools also runs an MCP server — read-only utility tools your AI agent can call directly. No key, no signup: requests are processed in memory and request contents are never stored.

Claude Code
claude mcp add --scope user --transport http clean-tools https://mcp.clean.tools/mcp
Any MCP client (streamable HTTP, authless)
https://mcp.clean.tools/mcp

Every tool, the REST API, and setup for Cursor, Claude Desktop, VS Code & more →