LiveCompose 2.0 Update Notes Brand New Interface Architecture Bottom Navigation Bar Reconstruction Three-Tab Layout: The layout includes Gallery, Compose, and Settings. The Compose page is now the default homepage. Independent Capture Button: The capture button sits on the far right of the navigation bar. Clicking this button opens the camera interface immediately. Returning from the camera brings you back to the Compose page automatically. Compose Tab Icon: The Compose tab now uses the App Logo as its icon. Full Adaptation for Dark and Light Modes Appearance Settings: The Settings page now supports a three-stage toggle among Follow System, Light, and Dark modes. Full Page Adaptation: All pages now support the theme adaptation. These pages include Gallery, Compose, Settings, Photo Details, and Export Preview. Text colors and backgrounds change automatically with the theme. Capture Page Exception: The capture page stays in Dark Mode to ensure a good camera preview experience. New Compose Page The Compose page serves as the default homepage. It shows the full picture of the LiveCompose project. Brand Identity: The page displays the App Logo and the brand slogan. Project Introduction: The text describes the project as an AI-driven, on-device, real-time composition system. Project Repository Cards: The section includes cards for LiveCapture (iOS App) and LiveCompose (Model Training Framework). Users can click these cards to jump to GitHub. Resource Links: The page provides links to the GitHub Organization and the HuggingFace Model Hub. These links include brand icons. Core Technology Display: The section highlights key technologies. These technologies include the Adacrop reinforcement learning model, gyroscope motion tracking, Apple Vision native detection, and multi-lens intelligent zoom. Core Developers Section: The section features JettyCoffee, ZyanNo1, and zzsyppt. It includes their avatars and GitHub links. Gallery Upgrades Multi-Select Deletion: Long-pressing a photo enters the selection mode. Users can select multiple photos and delete them with the trash icon in the top right corner. Guidance Banner: A banner displays the prompt: "Tap to browse · Long press to select · Enter to export beautiful cards". Enhanced Photo Details: The details page now shows the detection engine name, ISO, shutter speed, aperture value, and pixel dimensions. Browsing Experience: The browsing experience is now consistent with the native iOS Photos app. New Settings Page Composition Engine: The engine offers None, Fast, and Pro modes. Fast is the default mode. None: This mode uses the native Apple Vision framework and requires no extra models. Fast: This mode uses the lightweight Adacrop Student model to balance speed and accuracy. Pro: This mode uses the full Adacrop Teacher model to deliver the highest accuracy. Auto-Capture Switch: The camera triggers shooting automatically after aligning with the composition frame. Capture Delay: Users can choose among 0.5, 1.0, 1.5, and 2.0 seconds. Theme Mode Toggle: The toggle allows quick switching between theme modes. Share and Export System Save After Preview: Clicking export generates a card preview first. Users can save the card to the system album after confirmation. Beautiful Card Design: The card uses a 3:4 ratio and rounded corners. It includes a transparent App Logo, the shooting date, and full shooting parameters. These parameters include the model name, ISO, shutter speed, aperture, and resolution. Storage Change: Photos are now saved only inside the App. The system no longer writes photos to the system album automatically. Camera and Detection Camera Lifecycle Fix: The update fixes a race condition where the camera could not close when users entered and exited the capture page quickly. Coexistence of Three Composition Engines: The CropDetectionStrategy protocol unifies the detection interfaces for Vision and CoreML.

LiveCompose is dedicated to enabling every ordinary user to easily take professional-grade composed photos. Unlike the static rule of thirds lines of traditional cameras, we use AI to analyze the viewfinder in real time and combine it with the device's gyroscope to achieve physical-level tracking guidance. It actively "tells" the user how to move the phone to achieve the best composition and automatically takes the picture when the perfect composition is aligned.