Skip to content
Library Actively developed

comicmation

Find pages, page numbers, panels, speech balloons and their text in comic images — on-device Swift library + CLI

detected comic panels
detected comic panels
  1. v0.1.0 · Oct 2026 · Latest

    comicmation 0.1.0

    First tagged release, so apps can depend on a version instead of a local checkout:

    .package(url: "https://github.com/webmatze/comicmation", from: "0.1.0")
    

    Link the ComicmationKit product. It has no third-party dependencies; SwiftPM still resolves swift-argument-parser and Yams for the CLI target but does not build them into the app.

    What it does

    From a plain comic image: page count and page boxes (spreads are split), printed page numbers, panels in reading order, text blocks with their balloon outline and OCR text. Everything runs on-device with CoreGraphics and Apple Vision, in about 1 s per page (release) and 1–3 s in an app's debug build.

    Result format

    • schema_version 2, algorithm_version 1. Key cached results by both (AnalysisResult.currentSchemaVersion, AnalysisResult.currentAlgorithmVersion) and by the configuration.
    • New in this release: line_height and line_count per text block (#13).

    Since the first integration into Krazoom

    • Debug builds are about four times faster: 12.3 s → 2.5 s for a 2100×3150 page (#11).
    • Algorithm version for result caches (#12).
    • Line height per text block for zooming until the text is legible (#13).

    Not included

    The CLI is not attached as a binary yet: it is not signed or notarized (#5). Build it with make release. The optional Core ML panel model is not part of the package; make model builds it.