Skip to content

Getting started

Platform support

What runs today on macOS, Windows, and Linux, the current unsigned-build status, and system requirements — honest about tested versus in progress.

Tessera is developed and tested primarily on macOS (Apple Silicon). The indexing pillars run cross-platform, but Windows and Linux are still being brought to first-class parity. This page is honest about what works today.

Status by feature

CapabilitymacOSWindows / Linux
Tags, embeddings, search, NudeNetworksworks
Captions (Tier 2 VLM)works (mlx)swap the caption route to a local server
Faces — embedding & clusteringworksworks
Faces — detectorworks (Apple Vision)in progress (cross-platform detector)
Video ingestworksneeds ffmpeg / ffprobe on PATH
Desktop bundleworksin progress (per-OS build matrix)

Tags, embeddings, search, and NudeNet metadata work on all three platforms today through ONNX Runtime’s execution-provider probing — see Compute providers.

The remaining gaps are honest ones. Tier 2 captions use mlx, which is macOS-only; on Windows and Linux you point the caption route at any OpenAI-compatible local server. Face detection currently uses Apple Vision, so the on-Windows/Linux detector is still in progress; face embedding and clustering already work everywhere. Video ingest needs ffmpeg available on the path.

If you run from source rather than a desktop bundle, the indexing pillars work on all three platforms now.

Signing status

Builds are currently unsigned. This is expected until the first stable release; it is not a sign of compromise. Every release publishes SHA-256 checksums — verify your download before running it.

On macOS, Gatekeeper will refuse to open an unsigned build on a double-click. Use the right-click Open path (or System Settings → Privacy & Security → Open Anyway), documented step by step under Install. On Windows, SmartScreen may warn that the publisher is unrecognized; choose More info → Run anyway.

Code signing and notarization (Apple Developer ID on macOS) are planned, after which these steps go away.

System requirements

PlatformMinimum
macOS13 Ventura or newer, Apple silicon or Intel
Windows10 / 11, 64-bit (DirectX 12 GPU recommended)
Linuxx86_64, glibc 2.31 or newer

All platforms: 8 GB RAM (16 GB recommended) and roughly 10 GB of free disk for the app plus models. The first launch downloads about 3.5 GB of local models; after that Tessera runs fully offline. The catalog and thumbnail cache grow with your library.