Any image, anywhere
Photos, screenshots, memes, CSS backgrounds, canvases, video frames, inline SVG, or any element you pick on the page.
Free · Open source · Private by default
Every image
is now text.
MagicLens reads the words inside any image on the web and lays them back over it, exactly where they are. Select them, copy them, translate them, or ask about them, like any other text on the page.
Chrome · Edge · Firefox · Opera · Brave · Vivaldi · Arc
Reading text
This is live. Drag across the poster to select its text, then try Translate.
Video
Watch MagicLens work.
0servers, accounts or trackers
30OCR languages, read on-device
11AI providers, or none at all
0long tasks on the page while it reads
Features
Text you couldn't touch, now you can.
Pixel-perfect text layer
Every word sits on its own glyphs, stretched to their exact width. Copy gives you clean spaces and line breaks.
Local OCR, offline
Tesseract and PaddleOCR run inside your browser. No account, no upload, nothing leaves your device.
Translate in place
The translation is painted over the original paragraphs, like a printed page in your language.
Ask the image
Explain, summarize, or ask anything. Bring your own key, or run a small vision model locally with Ollama or LM Studio.
Smooth on busy pages
Every effect runs on the compositor. The page you're reading never stutters while MagicLens works.
How it works
Three moves. No learning curve.
-
1
Point
Hover an image and click the lens, right-click it, or press Alt+Shift+L.
-
2
Read
Local OCR finds every word. With an AI provider, Hybrid mode lets the model correct it while keeping exact positions.
-
3
Use it
Select, copy, translate in place, or ask a question. Esc puts the image back the way it was.
Privacy
It collects nothing.
Really, nothing.
There is no MagicLens server. Local OCR never touches the network. If you connect an AI provider, images go straight from your browser to the provider you picked, with your own key, only when you ask.
Read the privacy policy →- No account
- No analytics or telemetry
- No ads
- Keys stay on your device
- Open source, Apache-2.0
Bring your own AI
Works with the model you already use.
OpenAIClaudeGeminiOpenRouterGroqMistralxAITogether AIOllamaLM StudioAny OpenAI-compatible server
AI is optional. Local OCR reads, selects and copies without any of them.
Changelog
What changed, build by build.
0.2.117 Oct 2026
- 128px icon now follows the Chrome Web Store layout: 96×96 artwork with 16px transparent padding.
0.2.017 Oct 2026
- URLs, bare domains, email addresses and IPv4 addresses in the recognized text are now clickable links over the image.
- Hovering a link previews its destination; the panel reports how many links a read found.
- File names such as
main.tsorREADME.mdare not mistaken for addresses.
0.1.010 Oct 2026
- First packaged build: in-place text layer, local Tesseract and PaddleOCR, optional AI reading and in-place translation.
- Hover button, context menu, keyboard shortcuts and an element picker.
- Compositor-only effects, background updates, legal pages and store kit.
Install
Get MagicLens.
ChromeAlso Brave, Vivaldi, ArcComing soon
EdgeMicrosoft Edge Add-onsComing soon
FirefoxDesktop and AndroidComing soon
OperaOpera Add-onsComing soon
Can't wait? Build it from source in two commands.
Free, no ads, made with care.
If MagicLens saves you time, a coffee keeps it going.
Support on Ko-fi