fontloom

Cross-platform desktop font manager & previewer for Windows 10/11 and macOS — browse every font on your system (and loose font files on disk), preview them with your own sample text, compare typefaces side by side, tag/organize into collections, and get optional local-AI font‑pairing suggestions. Offline and privacy‑first: everything runs locally, no account and no cloud required.

Overview

fontloom scans your installed fonts plus any folders of loose font files (.ttf, .otf, .ttc, .woff, .woff2) and gives you a fast, searchable gallery. Type a custom sample string once and see it rendered across your whole library at any size, weight, and style. Pin fonts into a comparison tray, organize favorites into named collections, and export a specimen sheet. An optional local‑AI mode (via Ollama / llama.cpp‑compatible tiny models) suggests complementary heading/body pairings — but the core experience works fully offline with no AI.

Motivation

Designers, developers, and writers accumulate hundreds of fonts and have no good way to see them. The OS font pickers show a tiny one‑line preview, don't let you use your own sample text, and can't compare two faces at once. fontloom is a focused, native‑feeling desktop utility that makes choosing and organizing fonts fast — without uploading your font library to a web service.

Use cases

• Pick a font fast — render your headline/paragraph in every installed font at once and scan visually.
• Compare candidates — pin 2–6 fonts into a side‑by‑side tray with synchronized sample text.
• Organize a library — tag fonts (serif, display, mono, "brand", "avoid") and group into collections.
• Audit loose files — point fontloom at a downloads/fonts folder and preview un‑installed font files without installing them.
• Build a specimen sheet — export a PNG/PDF specimen of a font or a collection to share.
• Find a pairing — (optional local‑AI) get heading/body pairing suggestions from fonts you already own.

How to use

Windows 10/11 quickstart

1. Download the portable fontloom-win-x64.zip from Releases (or the MSIX installer) and run fontloom.exe.
2. On first launch fontloom enumerates installed fonts via the system font directory and DirectWrite.
3. (Optional) Add a folder of loose font files: File → Add font folder….
4. Type your sample text in the top bar; adjust size/weight with the sliders.

macOS quickstart

1. Download fontloom-macos.dmg from Releases, drag fontloom.app to Applications, and launch it.
2. fontloom enumerates fonts from the system and user font directories (~/Library/Fonts, /Library/Fonts, /System/Library/Fonts) via Core Text.
3. (Optional) Add a folder of loose font files: File → Add font folder….
4. Type your sample text and preview across your library.

Cross‑platform note: the core logic lives in a UI‑free Fontloom.Core library so the same font parsing, indexing, tagging, and specimen rendering runs on both platforms; only the shell (window + system font enumeration) is platform‑specific.

Example workflow

1. Launch fontloom → it indexes your installed fonts.
2. Set the sample text to The quick brown fox — 0123456789 and size to 32pt.
3. Filter to Sans‑serif + weight ≥ 600 using the sidebar facets.
4. Pin three candidates into the comparison tray.
5. Tag the winner brand/heading and add it to the Website 2026 collection.
6. Export → Specimen (PDF) for the collection to share with the team.

Local‑AI integration (optional)

fontloom can talk to a local, OpenAI‑compatible endpoint (Ollama or llama.cpp server) to suggest font pairings and generate short, human‑readable descriptions/tags for a typeface based on its metadata and metrics.

• Tiny‑model friendly: Llama 3.2 3B / Qwen2.5 3B / Phi‑3‑mini class for text pairing suggestions; MiniCPM‑V class if visual specimen analysis is enabled.
• Local only: requests go to http://localhost:11434 (or your configured endpoint). Nothing leaves your machine.
• Metadata only: the AI prompt includes only font metadata (family/subfamily, inferred classification, style flags, weight/width) — font binaries are never uploaded.
• Graceful fallback: a reachability probe runs first; if no model is available, the app automatically falls back to rule‑based pairing heuristics (classification + contrast rules).
• Off by default: enable it explicitly in Settings → Local AI and adjust the endpoint there when needed.

Current status / milestones

🚧 Bootstrapping. This repo currently contains the plan and backlog. Tracked milestones:

• M1 — Core & indexing: font file parsing + system enumeration + searchable index.
• M2 — Gallery UI: custom sample text, size/weight controls, virtualized grid.
• M3 — Organize: tags, collections, favorites, persisted store.
• M4 — Compare & specimen export: comparison tray + PNG/PDF specimen sheets.
• M5 — Local‑AI pairing (optional): endpoint probe + pairing suggestions with fallback.
• M6 — Packaging: Windows (portable zip + MSIX) and macOS (.app/.dmg) builds in CI.

See PLAN.md for scope, architecture, and non‑goals.