Open Source Google Lens Alternatives That Actually Work in 2026

Free isn’t the same as open source, and anyone searching for an open source Google Lens alternative usually knows the difference. A closed app with no price tag doesn’t answer the question.

So here’s the list I’d actually give someone who wants to de-Google this part of their life, organised by the job you need done and the device you’re on. I checked every licence and release date on GitHub, F-Droid or the relevant store on 16 September 2026. Open source projects move fast in both directions, so treat the dates as a snapshot.

First, the part that decides whether this article can help you at all.

What open source can and can’t replace

Google describes Lens as a way to “use your camera or an image to search”, and lists identifying “what plant is in your friend’s apartment, or what kind of dog you saw in the park” alongside finding similar clothes and translating text from over 100 languages.

That’s two very different jobs sharing one button:

• Reading text from a photo, screenshot or scan. Open source is genuinely strong here. The engines are mature, run locally and handle more than 100 languages.
• Recognising things: this plant, that lamp, this landmark. That depends on matching your photo against an enormous index of the web. I couldn’t find an open-source app that does this the way Lens does, and I’d be suspicious of any list that claims one.

If you mostly use Lens to copy text, an open source Google Lens alternative is realistic today. If you mostly use it to identify objects, skip to the reverse image search section, which is the closest thing and comes with a catch.

The engines everything else is built on

You rarely use these directly, but almost every app below is a front end to one of them. If you want the background on how recognition works, what OCR is and how it works covers it.

Tesseract is the default. Apache-2.0 licensed, it ships as a library and a command-line program, and its README is upfront that the project “does not include a GUI application”. It can recognise more than 100 languages out of the box and can be trained on others. Version 5.5.3 was released on 24 July 2026, so it’s very much alive.

PaddleOCR is the newer, deep-learning-heavy option from the PaddlePaddle project, also Apache-2.0. It’s a Python toolkit aimed at developers and document pipelines, advertises support for 100+ languages, and now includes an official browser SDK, PaddleOCR.js. The latest release, v3.7.0, is from 11 June 2026. If you’re not comfortable installing Python packages, this isn’t your starting point.

On the desktop: grab text off your screen

This is where Lens-style “copy that text” works best in open source.

NormCap (Windows, macOS, Linux) is the closest thing to a desktop Lens for text. You capture a region of the screen and get text back instead of an image. It uses Tesseract underneath, and the project site states that “detection is performed on your device and no telemetry data is collected.” Licence GPL-3.0. The last tagged release is v0.6.0 from 31 August 2025; the main branch was last committed to in July 2026, with dependency updates still being merged in September, so it’s quiet rather than abandoned. Installers are on GitHub, and on Linux the project recommends Flathub.

Frog (Linux, GNOME) is the GNOME-native take on the same job. It extracts text from anything you can capture, including video and web pages, and also decodes QR codes. MIT licensed, available on Flathub and Snap, and its Flatpak bundles Tesseract. Latest release 1.6.0 on 24 April 2025; the repository was last updated in May 2025. Fine to use, but it has gone quiet.

TextSnatcher (Linux) still appears in a lot of recommendations. GPL-3.0, built on Tesseract 4.x. Its last release, v2.0.0, dates from 15 March 2022, and the Flathub page labels the package as community-provided and “not verified by, affiliated with, or supported by” the developer. It may still work for you. I’d pick Frog or NormCap first.

If you’re on Windows and want more than open-source options, the sibling piece on Google Lens alternatives for PC covers the wider field.

For scanned PDFs: OCRmyPDF

A lot of Lens use is really “I have a scanned document and I need to search it”. For that, OCRmyPDF is a better fit than any camera app. It adds an OCR text layer to a scanned PDF so you can search and copy from it, produces PDF/A by default, and uses Tesseract, so the same 100+ languages apply.

It runs on Linux, Windows, macOS and FreeBSD, with Docker images for x64 and ARM. Licence MPL-2.0. The latest release, v17.11.0, came out on 28 August 2026. It’s a command-line tool, which is the price of admission, but the basic usage in its README is one line: ocrmypdf myfile.pdf myfile.pdf.

On Android: fewer good options than you’d think

Image Toolbox is the one I’d point to first. It’s an image editor with a large OCR feature rather than a dedicated scanner, which sounds like a compromise until you see the OCR section: 120+ languages, several Tesseract model sizes, batch recognition to a file, and creating searchable PDFs. Apache-2.0. The F-Droid build is the “foss” variant, which per the README replaces Google’s ML Kit with open libraries and removes analytics and Crashlytics. Version 4.1.0-foss was added to F-Droid on 3 September 2026; GitHub already had 4.2.0 on 4 September.

Text Fairy is the name that comes up in every degoogle thread, so it needs a straight answer. The public GitHub repository hasn’t had a commit since 12 November 2020, and its last GitHub release is from 17 November 2020. It isn’t on F-Droid. The Google Play listing was last updated on 12 May 2024 and is marked as containing ads. The repository includes an Apache 2.0 licence file. In practice that means you can’t assume the app you install from Play matches the open code, and the open code is nearly six years old.

On iPhone: open front ends, closed engine

This is the awkward one. I found exactly one open-source OCR app for iPhone worth naming, and it hasn’t been updated in more than three years.

That app is OpenFind, which finds text in your photo library and live through the camera, offline, with no data collected according to its App Store listing. Its GitHub repository carries an MIT licence. But its code was last updated on 10 February 2023, and when you look at the source, the recognition is done by Apple’s VNRecognizeTextRequest, part of the Vision framework.

That’s the pattern on Apple platforms. Apple’s text recognition is closed source; developers reach it through Vision, and VisionKit gives apps the standard Live Text interface. Open-source Mac tools like TRex (MIT, v2.0.0 released 13 February 2026) put a menu-bar screen grabber on top of the same framework. The app code is inspectable. The part that reads the text isn’t.

Which brings me to a disclosure. I build Textora, an iPhone OCR app. Textora is not open source. It uses Apple’s Vision framework on-device, works offline and needs no account, which covers some of what privacy-minded readers want, but it doesn’t meet the standard this article is about. If “I can read the code” is your requirement, don’t count it. There’s a broader look at privacy-focused options in Google Lens alternatives that respect your privacy.

The “what is this?” gap, and the one partial answer

Search by Image is an open-source browser extension (GPL-3.0; v8.5.4 released 16 June 2026) for Chrome, Firefox, Edge, Opera and Safari. You select an image on a page, capture part of the screen or upload a file, and it sends that to reverse image search engines you choose. Its supported list includes Google Lens, Bing, Yandex, Baidu, Kagi, TinEye and dozens of stock photo and trademark databases.

It’s useful, and it’s honest about what it is. But the matching still happens on those services, and nearly all of them are closed. You’d be using an open-source remote control for a closed search engine, and for the obvious engines, the image still goes to Google, Microsoft or Yandex. That’s not a failure of the project. Identifying arbitrary objects needs an index of the web, and nobody has built an open one at that scale.

The table

Checked on 16 September 2026. “Last release” is the latest non-prerelease on GitHub unless noted.

What I’d actually install

• Linux desktop: NormCap or Frog, plus OCRmyPDF for scans.
• Windows or Mac desktop: NormCap, plus OCRmyPDF if you deal with PDFs.
• Android: Image Toolbox from F-Droid.
• iPhone: there’s no fully open stack. OpenFind is the closest, with the caveats above.
• Identifying objects: no open source Google Lens alternative exists for this. Search by Image is the least-bad bridge, knowing where the image ends up.

I build Textora, an iPhone OCR app that is not open source, so I have an interest in this topic but no project in the list above. Licences, release dates, store listings and README details were checked on 16 September 2026 from the links below. Projects release, stall and restart all the time; if anything here is out of date, email support@textora.app and I’ll re-check and note what changed.

Sources: Google Lens · Tesseract on GitHub · PaddleOCR on GitHub · NormCap on GitHub · NormCap project site · Frog on GitHub · TextSnatcher on GitHub · TextSnatcher on Flathub · OCRmyPDF on GitHub · Image Toolbox on GitHub · Image Toolbox on F-Droid · Text Fairy on GitHub · Text Fairy on Google Play · OpenFind on GitHub · OpenFind on the App Store · TRex on GitHub · Search by Image on GitHub · Search by Image supported engines · Apple — VNRecognizeTextRequest · Apple — Enabling Live Text interactions with images