Privacy
A short page because there's not much to say.
Tiramisu does not collect telemetry, analytics, crash reports, or "anonymous usage data." There is no analytics SDK linked into the app. There is no server we phone home to.
By default — everything except the cloud Replicate backend.
VNGenerateForegroundInstanceMaskRequest, runs on the Neural Engine. Never leaves the machine.Three things, only when you opt in:
If you pick AI → Generative Fill Settings → Replicate and enter an API key, your image (cropped to the fill region) is sent to replicate.com for inference. Their privacy policy applies. We don't proxy or cache anything.
If you choose Local FLUX-Fill, the bootstrap script downloads the model weights (~24 GB) from huggingface.co using your Hugging Face token. After the download, no further network calls.
When Sparkle auto-updates land in v0.2+, the app will check for new releases on a schedule. This is a single GET to a static appcast.xml on tiramisu.hanley.world — no body, no telemetry, just "is there a newer version." Opt-out from Settings → Updates.
This site (tiramisu.hanley.world) is static HTML / CSS / SVG. No analytics, no fonts loaded from Google or Adobe, no third-party JS. Web server logs (timestamps, IPs, user-agents) exist for the few minutes nginx keeps them; they're not aggregated or shared.
Privacy questions: open an issue on GitHub or email privacy@hanley.world.
Last updated 2026-05-07 · This is the canonical statement. If anything changes, we'll update this page and note it in the release notes.