Read your local comic library with ease and translate on demand using your own AI.
Browse nested subdirectories and read your comic archives effortlessly. Whenever you hit a language barrier, connect your existing AI subscription or local LLM for context-aware translations that understand full scenes and character relationships—proofread overlays live, and finalize clean pages whenever you're ready.
Free and open source. Local speech bubble detection and OCR run on your PC with zero subscription lock-in.
Effortless reading meets context-aware translation.
No database lock-in, no rigid subscriptions. Browse your entire collection and bring your own AI when you need it.
Pick any comic folder
Recursively scans your directory tree instantly, surfacing every subfolder containing comic pages and remembering your reading progress across sessions.
Context-aware translation on demand
Translate dialogue using your own OpenAI-compatible API or local LLM. Multi-bubble context ensures natural conversational nuance across panels instead of stiff word-by-word substitutions.
Live overlays & quick in-place editor
Translations render as sharp vector overlays directly over speech bubbles. Catch missed text, tweak dialogue phrasing, or toggle original artwork instantly with the T key.
Finalize into clean image releases
Once proofreading is complete, render and rasterize the clean dialogue into permanent, high-resolution image files in a sibling folder with one click (Ctrl+B).
Engineered for collectors, readers, and translators.
Deep architectural control: background processing queues, switchable vision models, and reader ergonomics.
Asynchronous queue manager
Heavy bubble detection, OCR, and AI translation process in a dedicated background queue with rate and pause controls, keeping the reading UI completely fluid and bottleneck-free.
Quick in-place translation editor
Click and edit text directly inside speech bubbles. Correct OCR typos, refine tone, or manually translate missed bubbles right on the canvas before committing.
Switchable local ONNX models
Easily swap between different local ONNX bubble segmentation models to optimize detection accuracy for varied comic art styles, resolutions, and page layouts.
Dual translation modes
Choose Token-Efficient Mode (local ONNX + local OCR + text LLM for ultra-low cost) or Visual Hybrid Mode with vision LLMs for deep scene context and zero OCR typos.
Continuous nav with lock-in zoom
Reaching the last image of a chapter seamlessly transitions into the next folder's first page. Your chosen zoom level and fit mode remain locked across folder boundaries.
100% offline-first privacy
Neural speech bubble detection and Japanese OCR run strictly on your local hardware. Your files stay on your machine, with zero external telemetry.
One-click rasterization
Press Ctrl+B to finalize your polished overlays into clean, standalone image files in a sibling folder—ready for web viewers, tablets, or archive sharing.
Safe archive management
Batch-rename chapters with live diff tables, sort subdirectories, and delete unwanted folders safely via Windows Recycle Bin integration.
See KomaShift in action.
Reading view, page selection, live activity logs, and custom AI model configuration in native Windows dark mode.
Reading view
Folder browser, full comic page view, and thumbnail strip in a unified interface.
Page selection & caching
Select specific pages to translate or process entire chapters with cache tracking.
Real-time activity log
A timestamped trace of every background step: bubble segmentation, OCR, and AI requests.
BYO AI model configuration
Easily configure OpenAI-compatible endpoints, custom models, and token-efficient engines.
Questions, answered.
Everything about library browsing, translation models, privacy, and file safety.
01What makes KomaShift different from other comic readers?
KomaShift is built as a frictionless local comic reader first: point it to any comic folder, enjoy continuous chapter navigation with locked-in zoom, and browse nested subdirectories without importing. When you encounter untranslated chapters, bring your own AI key or local model to translate with full context, review the overlays live, and finalize clean pages at your own pace.02How does context-aware AI translation work?
Unlike legacy machine translation that processes isolated words or disconnected speech bubbles, KomaShift leverages Large Language Models (LLMs) to understand scene relationships and character dialogue across panels. Choose Token-Efficient Mode (local detection + OCR + text LLM for pennies per volume) or Visual Hybrid Mode (vision LLMs like Gemini or GPT-4o for nuanced artistic context).03Can I use my existing AI subscriptions or run local models?
Yes. KomaShift supports standard OpenAI-compatible endpoints, allowing you to use your existing API keys (OpenAI, Gemini, Anthropic, OpenRouter) or run 100% offline models via local tools like Ollama and LM Studio. You only pay provider rates for what you use, with zero middleman markups.04Why doesn't KomaShift automatically overwrite my original images?
Translation models and OCR can occasionally miss text or misinterpret phrasing. KomaShift keeps translations as an editable live draft overlay so you can proofread, correct typos, and adjust styling. Once you are satisfied, one-click rasterization (Ctrl+B) finalizes the clean, translated pages into a sibling folder, leaving your original files untouched.05How does the queue manager prevent UI freezes?
All compute-heavy operations—neural bubble segmentation, OCR, and AI network requests—are routed through an asynchronous background queue. This prevents UI stuttering, respects API rate limits, and gives you real-time visibility into token costs and page progress while you continue reading.06Which platforms are supported?
KomaShift is currently built as a native desktop application for 64-bit Windows 10 and Windows 11 (distributed as a standalone portable executable). Support for macOS, Linux, and Android is planned for future releases.
Read your comic library without language barriers.
Effortless folder navigation, context-aware AI translation on demand, and complete control before finalizing.