Frequently asked
If something's missing here, open an issue.
A free, open-source, AI-native image editor for macOS. Real layered editing (like Photoshop), with AI woven through the editing surface from day one — generative fill, expand, remove, replace, soon a natural-language intent bar.
Built for the channel-of-one shipping daily across YouTube, IG, TikTok, and the rest. Not for graphic designers serving big-brand briefs — for creators serving themselves.
Yes, actually free. No subscription, no Pro tier, no "free for personal use" trap. The editor, every AI integration, every export target — yours, forever.
The catch: it's pre-v0.1, in active development, and you bring your own AI compute (either an API key for Replicate, or a one-time setup of mflux for fully local generation). No surprise bills, but no hand-holding either.
No. Zero telemetry, zero analytics SDKs, zero "anonymous usage data." We don't have an analytics dashboard because we don't collect anything to put on one.
Network traffic only happens when you explicitly use the cloud Replicate backend (your image goes to Replicate's servers, per their privacy policy). Local FLUX-Fill runs 100% on-device. See Privacy.
Yes — pick the Local FLUX-Fill backend in AI → Generative Fill Settings. After a one-time ~24 GB model download, every generation runs on your Mac's Neural Engine. No data leaves the machine.
The setup is automated by ./scripts/bootstrap.sh from the source repo, or via the Help → Install Local FLUX-Fill… menu item in the app.
macOS 26.0 (Tahoe) or newer · Apple Silicon (M1 or newer) · ~30 GB free disk if you want Local FLUX-Fill (mostly the model weights).
Building from source also requires Xcode 17+ and xcodegen.
It's pre-v0.1, in active development. The app builds and runs; layered editing works; both AI backends generate; generative fill / expand / remove / replace all work. We haven't shipped a signed DMG yet — you build from source. A natural-language intent bar is planned for v0.2.
If you're a developer or early-adopter creator, dive in. If you want a daily driver, watch the roadmap and wait for the first DMG.
Pre-v0.1: clone and build from source. See Get started.
Once the first signed DMG ships: github.com/hanley-tech/tiramisu/releases/latest and (eventually) brew install --cask tiramisu.
No, and probably never. The Local FLUX-Fill backend spawns a subprocess at a user-installed path, which Mac App Store sandbox rules don't allow. Direct download via signed/notarized DMG (and Homebrew Cask) is the path forward.
Mac-only for now. The whole stack (Swift 6, AppKit, Apple's Vision framework, Core ML, Apple Silicon Neural Engine) is bet on macOS — that's where we get to be best-in-class. Cross-platform would require a different bet.
An iOS / iPadOS port is conceivable post-v1; we share the rendering pipeline. No promise yet.
Likely AGPL-3.0 — protects against a closed-source fork being resold as a SaaS. Final license lands in the LICENSE file at the repo root before v0.1. Until then, treat as "all rights reserved" pending license selection.
Please. Open an issue first to discuss bigger changes, then PR. Read CLAUDE.md at the repo root for code conventions (xcodegen-only, no in-tree edits to the .xcodeproj, AGPL-aware contributions).
Tiramisu has layers. Image editing has layers. Same idea, two crafts. Plus it sounds nicer than "FreeShop" or "Photoclone."