漫画与图文内容 AI 翻译工作流设计师

You are an expert in building automated manga/comic translation pipelines using AI. I want to create a workflow that translates visual content from [source language] to [target language].

Please design a complete pipeline:

1. Text Detection & Extraction

• OCR approach for manga-style text (vertical/horizontal, varied fonts, sound effects)
• Handling speech bubbles vs. narration boxes vs. onomatopoeia
• Recommended models/tools (PaddleOCR, Tesseract, manga-specific models)
• Output format: JSON with bounding boxes, text content, and text type

2. Translation Engine

• Context-aware translation that maintains:
  • Character name consistency across chapters
  • Tone/register per character (formal vs casual)
  • Cultural adaptation of jokes, idioms, and sound effects
• Glossary management system for recurring terms
• Prompt template for LLM-based translation with manga context

3. Image Editing & Typesetting

• Text removal/inpainting strategy for original text
• Font selection rules by text type (dialogue, narration, SFX)
• Text placement algorithm respecting bubble boundaries
• Handling right-to-left vs left-to-right reading direction

4. Quality Assurance

• Automated checks: missing translations, text overflow, readability
• Side-by-side comparison view generation
• Human review workflow integration

5. Batch Processing

• CLI command structure for processing entire chapters/volumes
• Progress tracking and error recovery
• Output formats (PNG, PDF, EPUB)

Provide the pipeline as a step-by-step guide with tool recommendations and example configs.