krites — Proposed Feature Specifications¶

This document outlines brief design specifications for proposed future capabilities that extend the core philosophy of krites: local-first, privacy-respecting, and highly deterministic.

1. "Perfect Group" Composite Proposals¶

Intent: Automates compositing in-app to ensure everyone's eyes are open in group bursts. Anticipated Functionality: Uses dedup and faceprovider to map faces in a burst. Selects the best base frame and uses the Inpainter interface to patch in open eyes from adjacent frames.

2. Lightroom AI Mask Pre-generation (via XMP)¶

Intent: Offload compute for generating Subject/Background masks from Lightroom to krites. Anticipated Functionality: Runs a local ONNX segmentation model against keepers. Generates paths for "Subject" and "Background" and writes them into Camera Raw XMP sidecars.

3. Voice-Driven Culling & Ergonomics¶

Intent: Alleviate ergonomic strain during culling via hands-free voice control. Anticipated Functionality: A continuously-listening local ONNX voice recognition model maps vocal commands ("Keep", "Reject", "Next") to Studio keyboard event triggers.

4. Zero-Infrastructure Client Proofing (WASM)¶

Intent: Provide a self-hosted, serverless client proofing alternative via WebAssembly. Anticipated Functionality: Compiles the engine to WASM and generates a standalone HTML/JS bundle with web-ready JPEGs. The client reviews in their browser and sends back a JSON file of selections.

5. Technical Feedback & Gear Analytics¶

Intent: Provide macro-level insights into gear performance and shooting habits. Anticipated Functionality: Cross-references EXIF metadata with QualityAnalyzer scores to aggregate data (e.g., motion blur rates per lens) into a visual dashboard in the Studio UI.

6. Narrative "Spread" Suggestions¶

Intent: Automate the initial scaffolding of album spreads by grouping cohesive photos. Anticipated Functionality: Evaluates color palettes and timestamps via AestheticScorer to group 2-5 visually cohesive frames together, tagging them for export.

7. Smart "Second Shooter" Camera Syncing¶

Intent: Solve chronological syncing of photos from multiple cameras with drifted clocks. Anticipated Functionality: Identifies visual overlaps (flash bursts, identical scenes) to calculate the exact millisecond time-drift offset between cameras, virtually aligning the timeline.

8. AI-Assisted Scene Organization & Naming¶

Intent: Deliver a client-friendly final product without manual folder sorting. Anticipated Functionality: Uses scene detection and timestamp clusters to categorize the day into chapters (e.g., "Ceremony") and dynamically renames files.

9. Emotional Arc / "Vibe" Sorting¶

Intent: Allow instantly locating photos with specific emotional tones. Anticipated Functionality: Extends FaceAnalyzer to tag frames with "vibe" metadata (e.g., "High Energy", "Intimate"). The UI provides filters to instantly pull exact moods.

10. VIP / Hero Tracking¶

Intent: Ensure the most important subjects are heavily represented in the final keep pile. Anticipated Functionality: The FaceAnalyzer runs facial similarity matching against a reference photo of the couple, automatically bumping the "Keep" score of any frame containing them.

11. Social Media Crop Proposals (9:16)¶

Intent: Auto-generate vertical slices for Instagram Stories/Reels. Anticipated Functionality: Uses saliency detection to generate a secondary set of 9:16 vertical crop coordinates from landscape photos, keeping the primary subjects perfectly centered.

12. Mixed-Lighting Consistency Warnings¶

Intent: Catch difficult-to-edit photos with conflicting color temperatures before Lightroom. Anticipated Functionality: The QualityAnalyzer detects scenes with extreme mixed lighting and flags them with a warning badge in the UI.

13. "Sneak Peek" Auto-Generator¶

Intent: Instantly generate a diverse preview gallery for the client the morning after. Anticipated Functionality: Algorithmically selects 10-15 of the highest-scoring keep photos that are narratively and visually diverse, auto-exporting them into a standalone folder.

14. Creator Look & Profile Marketplace¶

Intent: Allow photographers to share, import, and sell custom culling criteria and color grades. Anticipated Functionality: Implements an import/export mechanism for .krites-look and .krites-profile files.

15. Automated B-Roll / Cinemagraph Generation¶

Intent: Turn still photo bursts into highly engaging video assets for social media. Anticipated Functionality: Uses optical flow to align and stabilize high-motion burst clusters, exporting them as seamlessly looping GIFs or short MP4s.

16. Guest Coverage Matrix ("Did I Miss Anyone?")¶

Intent: Ensure no critical family members are accidentally excluded. Anticipated Functionality: Runs background face-clustering across the entire shoot to generate a visual dashboard of every unique face.

17. Faux-Bokeh / Smart Background Defocus¶

Intent: Salvage well-composed photos that suffer from distracting backgrounds. Anticipated Functionality: Uses a local depth-estimation model to artificially simulate a shallow depth of field (e.g., f/1.2 blur) exclusively on the background.

18. Automated Blog Post & Storyboard Builder¶

Intent: Eliminate the tedious manual layout work required to post a wedding. Anticipated Functionality: Generates a beautifully laid-out HTML/Markdown masonry grid of the top 50 photos, ready to be pasted into WordPress.

Competitor-Inspired Features¶

19. Automated Face Zoom Panels¶

Source Citation: Inspired by Narrative Select Intent: Eliminate the need to manually pan across a large group photo to check focus on every face. Anticipated Functionality: The Svelte UI leverages the bounding boxes from the FaceAnalyzer to instantly render a sidebar grid containing 100% zoomed-in crops of every detected face in the frame simultaneously.

20. Adaptive Subject Dodging & Burning¶

Source Citation: Inspired by ImagenAI Intent: Create a 3D pop by dynamically relighting the subject versus the background. Anticipated Functionality: Uses local segmentation masks to automatically apply exposure adjustments—brightening the subject (dodging) and darkening the background (burning)—directly during the export pipeline.

21. Hyper-Specific Retouching Models¶

Source Citation: Inspired by Evoto Intent: Automate the most tedious, pixel-level Photoshop retouching tasks. Anticipated Functionality: Integrates highly specialized ONNX models into the Inpainter interface strictly designed for "stray hair removal", "glasses glare removal", and "clothing wrinkle smoothing".

22. Metadata "Code Replacements"¶

Source Citation: Inspired by Photo Mechanic Intent: Massively speed up metadata captioning and IPTC tagging for photojournalistic event delivery. Anticipated Functionality: A configurable text-expansion map (e.g., typing \bg\ expands to "Bride & Groom") that injects the full string directly into the IPTC/XMP sidecar data during review.

Killer USPs (Unique Selling Propositions)¶

23. "True Moment Reconstruction" (Shoot-Specific LoRA)¶

Intent: Flawlessly correct blinks or awkward expressions during one-off, unrepeatable moments where no burst frames exist. Anticipated Functionality: Rather than using a generic generative AI model, krites dynamically trains a tiny, temporary LoRA on the hundreds of other photos of that specific subject from that day. It reconstructs their exact eyes/expression tailored to the specific lighting and environment.

24. Audio-Driven Emotional Slideshow Generation¶

Intent: Bridge the gap between photography and hybrid video by automating cinematic slideshows. Anticipated Functionality: Ingests an MP3 audio track (e.g., wedding vows). Uses a local Whisper model to transcribe and perform sentiment analysis. It sequences the most emotionally resonant photos to perfectly match the pauses, swells, and cadence of the spoken words, exporting an MP4.

25. The "Album Gap" Analyzer¶

Intent: Prevent the catastrophic realization that a critical narrative beat of the wedding was missed or accidentally rejected. Anticipated Functionality: Cross-references the "Keep" pile against a standard wedding checklist (Rings, Dress, Cake, First Kiss). It throws a warning if a narrative gap is detected prior to export.

26. Interactive "Second-Guess" (The Fatigue Fighter)¶

Intent: Act as a tireless pair-programmer to combat decision fatigue during massive culls. Anticipated Functionality: Quietly tracks when the user manually rejects photos that mathematically align with their historical 5-star criteria. At the end of the cull, it presents a single review screen asking, "You usually love shots like this. Were you just tired, or do you want to reconsider these 5 frames?"

27. "Focus Rescue" via Generative Upscaling¶

Intent: Salvage out-of-focus photos that are emotionally critical but technically unusable. Anticipated Functionality: Integrates a heavy, generative upscaling and sharpening model natively into the pipeline. Instead of automatically discarding a blurry photo of a crying grandparent, the user can hit a "Rescue" button to computationally reconstruct the lost focus.

Workflow & Business Management USPs¶

28. The "Time-to-Edit" Delivery Predictor¶

Intent: Help photographers manage client expectations and their own schedules. Anticipated Functionality: By analyzing historical editing pace and the current size of the keep pile, the Studio UI displays a real-time dashboard projecting how many hours of Lightroom editing remain and an estimated delivery date.

29. Sensor Dust Auto-Flagging¶

Intent: Catch sensor/lens dust issues before importing into Lightroom. Anticipated Functionality: Mathematically detects a persistent dark spot appearing in the exact same coordinates across hundreds of photos (especially those shot at f/8 or narrower). Alerts the photographer so they can prepare to use batch spot removal tools.

30. CRM / Shot-List Integration¶

Intent: Ensure no contracted "Must-Have" photo is ever missed during the cull. Anticipated Functionality: Ingests a text/JSON shot-list exported from a CRM (like HoneyBook). Acts as an interactive checklist in the Studio UI, requiring the photographer to explicitly tag a keeper to each requested shot (e.g., "Bride with Aunt Judy").

31. Smart Vendor Tagging & Routing¶

Intent: Automate the delivery of photos to demanding wedding vendors. Anticipated Functionality: Uses scene detection to identify specific vendor subjects (table settings, makeup, florals). Automatically copies these specific keepers into labeled "Vendor Delivery" folders during export, ready to be sent to the florist or makeup artist immediately.

32. The "Emotional Consistency" Balancer¶

Intent: Ensure the final gallery delivers a balanced, emotionally resonant narrative. Anticipated Functionality: Analyzes the keep pile as a holistic set. Flags the user if the distribution is too skewed (e.g., "Your gallery is 90% serious/formal. Consider reviving 10 candid/laughing shots from the 'Maybe' pile to balance the tone.")

33. Personalized Deep-Learning Aesthetic Classifier¶

Intent: Move beyond rigid mathematical thresholds (e.g., MinSharpness = 0.5) to a true machine-learning understanding of the photographer's unique artistic style. Anticipated Functionality: After collecting thousands of keep and reject decisions across multiple shoots, krites uses this vast, personalized dataset to locally train a lightweight classification model (or fine-tune a pre-trained AestheticScorer via ONNX). Instead of relying on manually tuned thresholds, the model inherently understands what Hailey considers a "5-star" photo based on her specific aesthetic, composition, and emotional preferences.