<PhatLam/>/docsVI

Introduction and what's new

What PixGamma is, its three principles, major changes by version and the documentation index.

PixGammaUpdated 23 Oct 2026
tổng quan0.2.1

PixGamma upscales and restores photos with AI on the Mac. A small, blurry or heavily compressed image comes back at 2x to 16x with clearer detail, processed entirely on the Apple Silicon GPU. PixGamma ships as a macOS app, a pixgamma command for batch work and a local HTTP API for scripts, websites and other apps.

PixGamma is the new name of Pixora (up to 0.2.1): old installs update to PixGamma on their own, and the pixora command keeps working.

Current version: 0.2.1 (2026-10-20) · macOS 14 or later, Apple Silicon · See the Changelog.

The new interface, Light and Dark: the photo fills the window, a dock for modes, a filmstrip below, split comparison at 100%The new interface, Light and Dark: the photo fills the window, a dock for modes, a filmstrip below, split comparison at 100%
The new interface, Light and Dark: the photo fills the window, a dock for modes, a filmstrip below, split comparison at 100%

The photo fills the window; a small dock at the bottom picks the mode, scale and Xem trước (preview); the filmstrip sits below and the tuning column on the right collapses. Fit shows the whole photo with the preview area outlined, 100% and 200% compare before/after split or side by side. Light, Dark or follow the system.

What it is for

  • Product and e-commerce photos: label, keycap and packaging text stays correct; materials get no invented texture.
  • Portraits and old family photos: clearer eyes, hair and skin; old scans get their grain cleaned up with the original tones kept.
  • Wildlife, nature, landscapes, architecture: fur, whiskers, feathers and carved detail hold up in large prints.
  • Illustration, anime, graphics: clean lines, flat color areas.
  • Screenshots and UI: clearer text, Vietnamese diacritics kept.
  • Developers: a CLI for whole folders, JSON reports and a local HTTP API.

Three principles

  • Photos never leave the Mac. Everything runs through Core ML on the machine. No account, no server, nothing uploaded.
  • No invented text. Keycaps, labels and banners are detected separately by Apple Vision (including Vietnamese with full diacritics) and rebuilt with a faithful model. Text too small to rebuild correctly is kept faithful instead of being redrawn as other letters.
  • Your photo stays your photo. Original color, tone and facial features are kept. Generated detail always has a strength slider.

What's new

Version Highlights
New on GitHub (unreleased) New interface, Light and Dark, export one photo or all · PixGamma Photo v2 is the default photo model: crisper fur, hair and fabric, text regions still use Photo v1 · Recover describes each photo with Apple Vision, picks its own resemblance and has a safety lock for text, maps and products · face restoration no longer colours black-and-white or sepia photos
0.2.1 (Oct 20) Text regions run through PixGamma Photo v1: keycap letters are sharper and straighter (OCR 91.1%), 3x faster · Recover no longer redraws keycaps and is no longer killed by macOS on large images · checked across every mode, 1x–16x and grayscale, CMYK, transparent, TIFF, WebP and HEIC inputs
0.2.0 (Oct 20) PixGamma Photo v1, our own photo model (PSNR 30.5 dB vs 29.0 dB for its base, no color drift) · DMG installer · auto-update from GitHub with SHA-256 checks · dedicated text model, Vietnamese text recognition · halo removal · diffusion-based Recover mode
0.1.0 (Oct 18) Three-pane SwiftUI app, five Core ML modes, face restoration, CLI and local API

Documentation

Getting started

Features

  • AI modes: Auto, Standard, High Fidelity, Art & CGI, Low Resolution, adjustments.
  • Text protection: how PixGamma finds, measures and rebuilds text, Vietnamese included.
  • Recover: skin, hair and fur regenerated with diffusion.
  • Own models & training: PixGamma Photo, benchmarks, how models are chosen.

Integration

Reference