A local-first AI image creation workbench. The desktop app ships with an embedded Companion service — it works right after installation, with no Node.js to install and no manual pairing; multi-provider credentials are managed on your own computer.
GPT-Image-Studio_0.4.1_aarch64.dmg · ~32.9 MB (Your system wasn't recognized, so the most common platform is recommended — switch on the cards below.)
No ads · Your data never leaves your machine · Installers are unsigned (standard for free open-source distribution, see the first-launch guide)
Download the dmg and drag the app into Applications. Unsigned apps need a one-time approval on first launch — see the step-by-step guide.
Download dmg ~32.9 MBIntel builds may arrive later
The installer is unsigned. On first run, click "More info → Run anyway" in the SmartScreen prompt and it works normally.
Download installer ~35.0 MBThe build pipeline is ready — this button unlocks automatically once the release is out
No origin checks to bypass: make the AppImage executable and run it, or install the deb directly.
The build pipeline is ready — the buttons unlock automatically once the release is out
Need an older version or checksums? Head to GitHub Releases to grab any release.
Files downloaded from the command line are not quarantined by macOS, so Gatekeeper never intercepts them. Copy the line below into Terminal — it downloads the app, installs it into Applications, and opens it right away, no manual approval needed:
curl -fsSL https://raw.githubusercontent.com/honlnk/gpt-image-studio/main/scripts/install-desktop.sh | shThe install script is fully auditable: scripts/install-desktop.sh
Manual install (browser download)
The installer is not signed with an Apple Developer certificate ($99/year — hardly worth it for this project), so Gatekeeper blocks the first launch. This is the normal warning for unsigned open-source software, not a virus — the entire app (source code) is open for audit. Approve it once and it behaves like any normal app.
Open the downloaded .dmg file and drag GPT Image Studio into the Applications folder.
Double-clicking the app shows "can't be opened because it cannot be checked for malicious software". Go to System Settings → Privacy & Security, scroll down to the Security section and click "Open Anyway" next to GPT Image Studio. (Since macOS 15 the old "right-click → Open" bypass no longer works.)
If you prefer the terminal, clearing the download quarantine attribute works the same as option 1:
xattr -cr /Applications/GPT\ Image\ Studio.appAfterwards, open it from Applications or Launchpad like any other app. The embedded Companion service starts and connects automatically — open Settings to configure your provider credentials.
The web app already does the job — the desktop app takes it further.
The embedded Companion service starts and connects automatically with the app — no Node.js, no global npm package, no copy-pasting connection keys. It shares its data directory with the CLI Companion, so credentials configured in one work in both.
Some proxy providers return only image URLs instead of base64, and browser direct mode can fail to fetch them under CDN CORS policies. The desktop app fetches images locally on the server side — no such problem.
Conversations, images, and provider credentials all live on your own computer (~/.gpt-image-studio) and never pass through any third-party server. The app itself makes no tracking requests and shows no ads.
Built on the Tauri v2 system webview, the installer is only a few MB (versus hundreds for Electron). Its own window and separate data partition keep it fully independent from the web app in your browser.
This is the normal warning for unsigned open-source software, not a virus. Two ways to approve it: System Settings → Privacy & Security → "Open Anyway"; or run xattr -cr /Applications/GPT\ Image\ Studio.app in Terminal, then open the app normally.
Completely free, open source under the MIT license. No ads, no data collection — image generation calls the provider APIs you configure yourself.
No. Desktop conversations and images live in a separate app data partition and don't affect the web app in your browser; both do share the Companion provider credentials on this machine (~/.gpt-image-studio).
The three-platform build pipeline is ready (GitHub Actions produces the packages automatically). Windows (NSIS installer) and Linux (AppImage / deb) will ship with the next desktop release, and the buttons on this page will unlock automatically then.
The desktop app embeds the Companion local service: it works right after installation with no manual pairing, multi-provider adaptation and credential management all happen on your machine, and it isn't subject to browser CORS limits — some providers only return image URLs, which the web app's direct mode can't always fetch automatically. The desktop app has no such problem.