Manga Studio 0.2.0

The first desktop release: Manga Studio as a Tauri app for Apple Silicon Macs, signed with a
Developer ID and notarized by Apple.

Desktop app

• Tauri shell that starts the Python backend with the app and stops its whole process group on
  restart and quit.
• Onboarding in a small borderless window: checks the Mac, downloads Qwen-Image-2.1 Turbo if it
  isn't cached (about 32 GB, once), connects an AI and creates a first project.
• Library, settings, pipeline view and export sheet.

Studio

• Page layouts (right-to-left or left-to-right grids, diagonal gutters) or a vertical webtoon scroll,
  with a layout editor and reader.
• Panel layers: background, art, screentone and FX, and an SVG lettering layer.
• Vector lettering: speech, shout, thought and caption bubbles, horizontal or vertical text.
• Cast studio: each character's face, hair, outfit, palette, reference images and identity seed.
• Panel copilot for text to image, image to image and inpainting, with the cast's reference sheets.
• Inspector and versions viewer; reorder panels from the rail.

Pipeline

• Six stages from story text to lettered pages: analysis, script, layout, artwork, lettering, QA.
• Auto-lettering places bubbles in low-detail areas away from faces, and never overrides a bubble
  you moved.
• QA retries a panel up to 3 times, then flags it for a manual fix.
• A coverage map ties every span of the source to panels.
• Language stages run on your agent, a local Ollama model or a no-LLM heuristic.

Agents

• MCP bridge (studio mcp) with 26 tools for Claude Code, Codex or any MCP client.

Export

• Print PDF (B5 or A5 with bleed and margins, or digital), layered SVG per page, PNG pages or
  webtoon slices, and mp4 video with slow zooms and crossfades.

Fixes

• Story slugs that escape stories/ are rejected (a /files/%2E%2E/... request served the repo).
• story.yaml writes keep comments and quote strings PyYAML would retype.
• Debounced saves flush on unmount and on leaving a view instead of being dropped.

Requirements

• An Apple Silicon Mac with macOS 14 or later; 24 GB of memory recommended.
• uv installed (the app runs its Python backend with it).