Tracking the evolution of Aegis Photo Voyager. New features, enhancements, and performance optimizations delivered every two weeks.
A new bulk-cleanup tool under Tools → Photo Tools. Find unwanted photos by minimum image dimensions, minimum file size, AI quality score, or capture date range, combined with either "match any" (OR) or "match all" (AND) logic.
Library Manager criteria update a live summary of the impacted photo count, earliest and latest capture dates, and estimated disk space, with a paginated thumbnail table so you can review and uncheck individual photos before removing them.
Photo file sizes are now recorded as photos are discovered, with a background scan that backfills sizes for existing libraries so file-size filtering and cleanup work immediately.
Added an in-process CPU engine that loads GGUF vision models directly, with no server to run. The LLM Setup Wizard now defaults to it on machines without a discrete GPU, alongside Ollama, LM Studio and vLLM.
The map view migrated to OpenStreetMap basemaps, removing the "API key required" watermark, and now loads its data in the background so the view opens immediately.
Added an OpenGL visual warmup that prevents flicker when loading views, and resolved an application exit hang and socket teardown errors on Windows.
Added a dedicated Find Similar action to the selection toolbar. With several photos selected it targets the first one chosen in chronological order, with context-aware tooltips.
Pagination now switches between numbered page buttons in wide views and compact icon navigation when the window is narrow or the side detail panel is expanded.
A specialized utility to detect, sort, clean and deduplicate imported Apple iPhone media libraries, with improved wrapping for long file paths.
Built a complete Windows MSIX packaging pipeline with a native launcher, automated build scripts, store asset generation and signing workflows.
Local LLM request timeouts now scale to your GPU availability and hardware profile, with automatic repair of malformed JSON responses during structured extraction.
Face recognition gained multi-threaded background workers, incremental scanning, cluster conflict handling and faster vector lookups.
A dedicated quality analysis dialog and worker covering score computation, attribute breakdowns (sharpness, noise, focus, composition, contrast), caching, star synchronization and soft delete.
Overhauled test orchestration across 44 suites and 273 unit tests with offscreen platform configuration, headless WebEngine and multimedia mocks, and clean thread teardown.
A toggleable, non-modal right-side panel for inspecting photo metadata, AI summaries, EXIF attributes, ratings, tags and collection membership without leaving the grid.
The side panel switches to a two-column layout past 500px, caps landscape previews at 350px, and replaces the star button with an explicit 0–5 star dropdown.
Added zoom and pan controls, quick map centering, external editing in the companion photo editor, spoken narration and keyboard shortcuts to the panel preview.
The panel advances to the next photo when one is removed, and gained a maximize/restore toggle plus a resizable splitter handle.
Selecting a photo group thumbnail or cluster marker on the map now shows the first photo of that group in the side detail panel.
A new module for grouping photos into logical travel journeys, with its own database schema and improved background worker status tracking.
A Collection Details dialog with a date distribution histogram, statistics, GPS distance tracking and a chip-based layout for AI-driven insights.
The slideshow player gained storytelling modes, subtitle rendering of photo narrations, playback controls and additional transition effects.
Imports from external devices now use persistent file metadata caching and synchronization history to avoid rescanning media you already have.
Added support for custom Piper text-to-speech voices, audio padding adjustments and a unified voice management interface.
A "Welcome to version x.y.z" dialog highlights each month's release notes, with an Appearance setting to restore it if you dismiss it.
Optimized SQL joins and added graceful handling of offline files with thumbnail fallbacks and clear alerts instead of broken entries.
Imports can now be attributed to a contributor or device, and a Curation Staging Queue lets you stage, preview, visually deduplicate and sort new imports before committing them to your library.
Regenerate AI metadata for a single photo directly from the detail view, with a prompt editor for fine-tuning the instructions sent to your local model.
Compile up to 100 photos from a collection into a single high-resolution PNG — landscape or portrait, Clean Grid or Scattered Polaroid, with a customizable background.
A Sugiyama-framework layout algorithm aligns generations, groups couples and minimizes crossing lines on the family tree canvas.
Added zoom, fit and pan buttons directly on the tree canvas, automatic edge-scrolling while dragging a portrait, and a tidy grid for people not yet connected to anyone.
The Person Photos dialog now shows instant full-resolution popup previews on thumbnail hover, matching the main library view.
Media resolution, format and file size are now surfaced in the main interface, and the default local vision model was updated.
Added an ffmpeg fallback for video thumbnail and frame extraction when the standard image utilities cannot read a file.
Database queries and statistics in the Actions dialog now load per tab, on demand, with wait cursors during long operations.
Upgraded the date chooser with a premium, high-precision TimeSpinnerWidget. Users can now drag-scrub or directly type into year, month, day, hour, and minute fields.
Implemented automatic chronological swapping for "From" and "To" filters. If a user selects an end date earlier than the start date, the application intelligently swaps them to maintain a valid range.
Date filters now consistently include and preserve hour/minute precision, with the "Enable Precise Time Entry" toggle removed in favor of a always-available high-fidelity interface.
Added dedicated mount scripts and enhanced scan progress reporting for iOS devices, including file-by-file metadata tracking during import.
Replaced the legacy binary star toggle with a multi-level rating system, allowing for more nuanced organization of photo libraries.
Added password protection, custom site descriptions, and purpose fields to the static website generator for improved privacy and context.
Introduced a specialized AI-powered tool for technical quality assessment, providing overall scores and attribute breakdowns for every photo.
Added a streamlined wizard to the sidebar to help users quickly generate and customize shared photo galleries.
Implemented a database-backed activity log that tracks all background AI jobs, providing a persistent history of imports, assessments, and website generation.
Split AI settings into independent "Metadata" and "Quality" tabs, allowing users to configure different models (e.g., GPT-4o for metadata vs. Ollama for local quality checks).
Added support for user-defined LLM prompts specifically for technical quality assessment, enabling tailored AI analysis criteria.
Refactored image/video handling to enforce consistent exclusion of deleted and duplicate photos across all analytics and UI views.
AI-generated quality assessments now include mandatory reasoning notes, stored in the database for future reference and transparency.
Redesigned the thumbnail overlays with modern, circular rating badges that adjust dynamically to the application theme.
Added configurable Ollama context window sizing to optimize local AI performance on varied hardware.
Improved the Timeline Tunnel's "seek" performance and synchronization with manual filter inputs.
Consolidated the build orchestration scripts for Linux and macOS, ensuring environment consistency across DEB, SNAP, and DMG packages.
Rolled out a comprehensive dynamic theming system supporting seamless switching between Light and Dark modes across all GUI components, including the Timeline Rail and Detail View.
Major overhaul of the onboarding experience with provider-specific checks, vision/embedding model selection, and real-time connectivity validation.
Extended theme support to the Photo Journey and Cluster views, using JavaScript injection and CSS variables to ensure visual consistency in web-based views.
Optimized application startup time by implementing lazy loading for computer vision dependencies, preventing delays on systems without hardware acceleration.
Introduced icon and metadata caching for the People View, significantly improving responsiveness when browsing large libraries of identified faces.
Finalized automated build and deployment pipelines for Windows (Installer), macOS (DMG), and Linux (DEB/SNAP), including dlib bundling and virtual environment isolation.
Implemented a startup watchdog and improved error handling for the main detail view to prevent hangs during heavy metadata extraction.
Added a dedicated slideshow mode with smooth transitions and theme-aware controls for presenting photo journeys.
Migrated legal documents, licenses, and assets into the application bundle for improved portable execution and compliance.
Launched the "Smart Start" dialog, a comprehensive onboarding experience that guides users through AI configuration, library indexing, and initial system checks.
Introduced the Actions Dialog to centralize background task management and provide a unified interface for manual tool execution.
Implemented persistent geolocation caching to speed up place-based filtering and reduce external API calls.
Introduced a Cythonized security module and updated license verification logic to protect application integrity.
Enhanced desktop presence with high-resolution icons, desktop file association, and standardized application naming.
Established the foundational multi-threaded GUI architecture using PySide6, featuring a responsive Sidebar, Timeline Rail, and asynchronous Image Grid.
Built the first Windows-specific deployment flow featuring a PowerShell installer for face recognition dependencies and virtual environment awareness.
Implemented automated Snap package generation for Linux, ensuring easy distribution across various distributions.
Developed the core image loading and thumbnail generation engine with support for HEIF/HEIC and basic metadata extraction.
Download the latest version of Aegis Photo Voyager and start organizing your memories with precision and privacy.