batunii/chika

Chika is an Android app for reading comic books, designed around the idea that reading a full comic page on a phone screen is uncomfortable. Instead of making you pinch and zoom your way around a page, Chika automatically detects the individual panels and walks you through them one at a time. Tap the right side of the screen to step into the next panel, tap the left to go back, swipe to turn full pages. The panel detection runs entirely on the device using a small AI model trained on manga and comic data. No internet connection is needed once the app is installed. The model identifies where each panel and speech balloon sits on a page, then a second step orders those panels correctly for the reading direction and handles cases where panels are very small (grouping them together) or very large (splitting them at a sensible cut point). If the model finds no panels on a page, the app falls back to showing the full page. Supported formats are CBZ and CBR files, including archives using the newer RAR5 format. You import comics through the standard Android file picker, and the app copies them into its own storage. It saves your reading position per comic at both the page and panel level, so you can pick up exactly where you left off. The library screen shows cover thumbnails and lets you remove comics with a long press. The code is split into two modules. A platform-independent core module written in Kotlin handles the comic archive logic, panel ordering, and reading math, and is structured so it can also be used in a future iOS app. The Android-specific module handles file formats, the AI model, the Room database for library state, and the Jetpack Compose interface. Chika is licensed under the Mozilla Public License 2.0. The bundled panel-detection model is based on a YOLO model trained on the Manga109-s dataset. Building requires Android Studio, with a minimum target of Android 8.0.