Your photo never leaves your device.

This page explains what happens to your photo in Social9 and how you can check it yourself. The code that handles your photos is open source.

View the code on GitHubCheck it yourself

Apache-2.0 license · Includes tests · Every AI model is verified

Where your photo goes

Your photo goes from your device into your browser. When you are done, you save it or share it wherever you like. It is never sent to our servers.

Four steps, all on your device

Each step is explained in simple terms. Open the technical details if you want to know more.

  1. Step 1

    You open a photo

    Your browser reads the photo from your phone or computer. It is not uploaded anywhere. Social9 has no upload feature at all.

    • The browser decodes the photo itself, including iPhone HEIC photos.
    • We read the location, date and camera details only to show you what will be removed when you save.
    • Source code: src/image/load.ts
  2. Step 2

    The AI tools download once

    The first time you remove a background or blur faces, your browser downloads the AI model it needs. The model stays on your device, so next time it starts straight away, even offline.

    • Face detection: 233 KB (MIT license)
    • Background removal: 6.6 MB (Apache-2.0 license)
    • Object cut-out: 45.9 MB (Apache-2.0 license)
    • Upscaling: 4.9 MB (BSD-3-Clause license)
    • Object removal: 28.1 MB (MIT license)
    • Before a model runs, the app checks its SHA-256 hash against the one published in our code. If the file has changed, the app won't use it.
    • Source code: src/ml/models.ts
  3. Step 3

    Your edits run on your device

    Background removal, face blur, stickers and resizing all run on your phone or computer. When your device has a capable graphics chip, the AI uses it to work faster.

    • The AI runs with ONNX Runtime Web, using WebGPU or WebAssembly.
    • MODNet cuts out people, ISNet cuts out objects, pets and groups, and YuNet finds faces. Real-ESRGAN makes photos bigger and sharper, and MI-GAN removes objects.
    • Source code: src/ml/matting.ts and src/ml/faces.ts
  4. Step 4

    You save a new file

    When you save, your browser creates a new image file from the edited picture. Nothing is copied over from the original file, so the hidden location and camera details are left out.

    • JPG, PNG, WebP and AVIF files are created by your browser. GIF, BMP, TIFF, ICO, PDF and SVG files are created by our open-source code.
    • To hit an exact file size, such as 50 KB for a form, the app adjusts the quality and then adds a few empty bytes that image viewers ignore.
    • Source code: src/convert/encoders.ts

What your browser downloads

Social9 downloads files from three places. None of them receive your photos.

Social9The app itself and three small AI models No photos sent
jsDelivrThe open-source program that runs the AI models No photos sent
Hugging FaceTwo open-source AI models, for cutting out and removing objects No photos sent
Your browser enforces this listEach page gives your browser this list. If the page tried to connect to any other website, your browser would block it.

Check it yourself

Here are three ways to confirm that your photos stay on your device. The first one takes a minute.

Turn off your internet

Use each tool once so the AI models are saved. Then turn on airplane mode and edit a photo. Everything still works, which shows nothing needs to be sent.

Watch the network

Open your browser's developer tools and go to the Network tab. Edit and save a photo. You will only see files being downloaded, never your photo being sent.

Check the security policy

Each page tells your browser which websites it is allowed to connect to. Your browser blocks every other website. You can read this list in your browser's developer tools.

Read the code that handles your photos

The code that handles your photos is public on GitHub. It covers the AI, opening and saving photos, removing hidden details and the security policy. You can run the tests and check every AI model with one command.

View on GitHub

Try it yourself

cd core
npm install
npm test
npm run verify-models

The last command downloads every AI model and checks that it matches the published hash.

Read the source

The photo-processing engine, website and editor are in our public repository. The same security policy applies to the whole site.

Photo-processing engine

  • Downloading the AI models, checking them and running them
  • Opening photos and reading their hidden details
  • Creating saved files and removing hidden details
  • The security policy used on every page
  • Tests, and a script that checks every AI model

Website and editor

  • The website and editor interface
  • Illustrations and marketing pages
  • Backgrounds, stickers and patterns

The security policy covers this code too, so it can't send your photos anywhere.

Questions

Answers to the questions people ask us most about privacy.

  • The full website and editor are available in our public GitHub repository. The core folder contains the photo-processing engine and security policy. The same policy applies to every part of the website.

  • You can check three things on the live website today. The AI models must match the hashes in our code. The AI program must match its published hash. And the security policy your browser receives must match the one in our code. We can't yet prove that the rest of the website's code matches byte for byte. We are working on it.

  • No. We don't use analytics, ads or tracking cookies. The security policy would also stop a tracking script from sending data.

  • It's gone. Your photo was only ever in your browser's memory. The AI models stay on your device so your next edit starts faster. You can delete them at any time in your browser's site settings.

  • Please email sameer.fyi@hotmail.com. The security policy in our code repository explains how we handle reports.