A developer just dropped SnapName on Hacker News—a macOS utility that watches your screenshot folder and renames files using a locally-bundled Gemma 4 multimodal model. The twist? That bundled AI comes at a cost: the app's DMG installer weighs in at 5.3 GB, which is... substantial for a screenshot tool. But as the creator notes, that's actually pretty reasonable for the kind of power you're getting with a free local model that never phones home.

How SnapName Actually Works

The app runs on llama.cpp under the hood, bundling Google's Gemma E4B model directly into the application package. It watches either your macOS default screenshot location (Automatic mode) or any custom folder you point it at (Custom mode). When a new PNG, JPEG, HEIC, TIFF, or AVIF file lands in that watched directory, SnapName analyzes the image content and serves up three AI-generated name suggestions. Users can manually approve a suggestion or enable auto-save to let the first recommendation stick without any interaction required.

Privacy by Design (Not Just Marketing)

Here's where it gets interesting for enterprise and privacy-conscious users: screenshot content is processed entirely on your Mac. No API calls, no cloud inference, no data leaving your machine. The Gemma model files ship inside the app bundle itself, meaning everything runs locally via llama.cpp's runtime. For developers or anyone handling sensitive screenshots—code snippets, financial dashboards, internal docs—this approach eliminates an entire category of data leakage risk that cloud-based alternatives introduce.

CPU vs GPU: Real-World Performance

The developer reports SnapName runs fine on CPU-only Macs but also taps Apple Silicon GPUs when available. For M-series chips, this means the Neural Engine can accelerate inference significantly compared to running purely on performance cores. The app targets macOS 13 (Ventura) or later, which covers essentially all actively-supported Macs at this point. Whether you're on an older Intel machine or a shiny new M4 MacBook Pro, you get local AI naming—just with different latency profiles.

The File Size Reality Check

Yes, 5.3 GB for a screenshot renamer sounds insane until you remember what's actually bundled: the full Gemma E4B model plus llama.cpp runtime libraries for macOS (including GPU acceleration code). This isn't bloat—it's the cost of true offline AI. Compare that to apps that require subscription API calls or upload screenshots to third-party servers for processing. SnapName pays the storage price once at install time and never touches an external service afterward.

Ecosystem Integration

SnapName isn't trying to replace your capture workflow—it augments it. The app works alongside CleanShot, Shottr, macOS native screenshot tools, or literally any utility that saves images to a watched folder. This design philosophy means you don't have to switch apps or change how you work; just point SnapName at wherever your screenshots land and let it suggest better filenames from there.

Key Takeaways

  • Bundled Gemma E4B model makes for a 5.3 GB install but enables fully offline AI naming
  • llama.cpp runtime provides CPU and Apple Silicon GPU acceleration on macOS 13+
  • No screenshots leave your Mac—complete privacy by architecture, not policy
  • Supports PNG, JPEG, HEIC, TIFF, AVIF with three name suggestions per image
  • Works alongside existing screenshot tools via folder watching

The Bottom Line

SnapName represents the kind of project that makes you wonder why it didn't exist sooner—local-first AI that respects both your privacy and your workflow. The 5.3 GB footprint is a feature, not a bug: it's proof positive that no cloud dependency exists. For developers tired of screenshot folders full of "Screenshot 2026-05-30 at 4.42 PM.png" filenames, this is exactly the tool the ecosystem has been waiting for.