Skip to content

Latest commit

 

History

History
95 lines (66 loc) · 5.87 KB

File metadata and controls

95 lines (66 loc) · 5.87 KB

QLens

QLens is an image viewer — and more importantly, a universal image-tagging protocol designed for the AI era.

English · 中文

See CHANGELOG.md for version history.

QLens is built around an open image-tagging protocol (qltag.db) that any software, script, or AGENT can read and write:

Component Description
QLens QuickView Native Win32 + D3D11 ultra-fast viewer — instant start, HDR rendering, WIC full-format + decoder plugins
QLens Manager Qt file-manager style — thumbnail browsing, tag management (assign/color/combo filter), QC inspection (auto-detect overexposure/blur/color-cast)
QLens MCP Server Opens your image library to AI clients (Claude / Cursor, etc.) — search/tag/statistics/batch analysis

Why QLens?

Starting point 1: there's no "Picasa-like" viewer anymore. Picasa was light, fast, clean — double-click and it shows, no distracting UI. But it's long discontinued. ACDSee and the like are feature-complete but heavy and slow to start; other viewers are either light with poor browsing experience, or look good but aren't light. QLens aims to bring back that light, clean viewing experience.

Starting point 2: every viewer is missing something. All viewers can open, zoom, and page through images — fast and pretty — but none of them can answer:

"Of the 200 photos I took at the beach last year, which ones have closed eyes?"

QLens = viewing + an open tagging protocol, so your library can be searched and understood by AI.

┌─────────────────────────────────────────────────────────┐
│                  QLens Tag Protocol (qltag.db)          │
│   One SQLite DB per folder — readable/writable by any   │
│   software or AGENT                                     │
└───────────┬──────────────────────────┬──────────────────┘
            │                          │
   ┌────────▼────────┐       ┌─────────▼─────────┐
   │  QLens Manager  │       │ QLens QuickView   │
   │  (Qt file mgr)  │       │ (Win32 + D3D11)   │
   │  thumbs/tags/QC │       │ fast viewing + HDR │
   └────────┬────────┘       └─────────┬─────────┘
            │                          │
   ┌────────▼──────────────────────────▼─────────┐
   │            QLens MCP Server                  │
   │  opens the library to AI clients             │
   │  (Claude / Cursor …)                         │
   └──────────────────────────────────────────────┘

Core: Tag Protocol

Everything revolves around qltag.db — one SQLite database per folder, keyed by pure filenames, naturally tree-shaped. Tags are classified by how they are detected:

  • QC tags (qc): reliably detectable without AI (overexposure / blur / color-cast) — one-click local CV detection
  • AI tags (ai): require AI detection (red-eye / closed eyes) — assigned by external AGENTs via MCP
  • Regular tags: manual or arbitrary

See the Tag Protocol Spec

Quick Start

bin/qlens_quickview.exe  double-click or drag an image to view (F=100%, S=fit window, wheel=paging)
bin/qlens_manager.exe    browse folders, tag images, run QC detection (double-click image to view)

Typical workflow: view in QuickView → double-click to open Manager → select images and tag them → hit "QC detection" to batch-apply QC tags → use combo/QC filters to find images → (optional) let AI read the library via MCP to add AI tags.

System requirements: Windows 10 1809+ (HDR features need an HDR display; HEIC/AVIF etc. need WIC extension or decoder plugin).

HDR Test: Your Monitor's Real Brightness

Displays labeled HDR400/HDR600 usually deliver far less than the label (e.g. HDR400 often measures only 200–350 nits). Use QLens to find the real value:

  1. Open testdata/hdr/hdr_range_test.jxr and press F for 100% view
  2. The image shows 9 brightness blocks: SDR 100 / 200 / HDR400 / HDR600 / HDR1000 / HDR1400 / HDR2000 / 4000 / 10000 nits, each with a linear gradient and labeled peak
  3. The block where gray levels stop getting brighter (clamped) is your monitor's actual HDR peak

QLens renders true HDR images (16-bit+) with 16F physical passthrough (scRGB 1.0 = 80 nits) — no adaptive brightening/dimming, so what you see is the pixel's true physical luminance. This makes the image both a brightness meter and the most honest test of a monitor's HDR capability.

Before testing: enable HDR in Windows, set monitor brightness to max, disable eye-care/power-saving modes, and confirm HDR is active in Windows display settings.

Documentation

License

LICENSE