Read web manga in Safari

Recognition, erasing and typesetting happen on your iPhone or iPad; translation uses the LLM you choose.

On-device recognition

Manga text and bubble detection, Japanese manga-ocr and LaMa inpainting run locally with Core ML, on the CPU, GPU or Neural Engine — whichever measures fastest on your device.

Your own model

Works with OpenAI, Anthropic, DeepSeek, OpenRouter and any OpenAI-compatible service. Only the recognized text is sent to the service you configure.

Typeset into the bubble

Translations are fitted to each bubble, horizontally or vertically like the original; one swipe shows the original again.

Safari extension

Translate the whole page from the page menu, swipe left on one image, or let the floating ball translate images as you scroll.

Never twice

The same image — even a resized or re-compressed copy — is translated once, and repeated sentences reuse earlier translations.

Sync

Settings and, optionally, the translation cache sync through iCloud or your own WebDAV server; API keys are encrypted before upload.

Requirements

iOS / iPadOS 17 or later with Safari.
An API key for an LLM service (e.g. DeepSeek, OpenAI, Anthropic, OpenRouter) or your own OpenAI-compatible server.
About 330 MB of on-device models, downloaded from Hugging Face on first use and removable at any time.

Privacy

The developer collects no data; there are no accounts, analytics or ads. See the privacy policy.

New here? Read the documentation. Need help? Visit support.