Translation Modes

Two display modes control how the translation sits on the page. The backend you pick also affects how the final pixels are produced.

Display modes (Display tab)

The Display mode selector in the popup has two options:

Overlay (default)

Translated text is rendered in bubbles that sit directly over the original image. The original manga art stays completely untouched underneath, so you can always hover or click to see the original.

• Font size auto-adjusts to fit each bubble
• Supports hover-to-see-original and click-to-toggle
• Fastest visual result — no full-image redraw

Side-by-Side

Creates a flex wrapper with the original image and the translated version side by side, each at 50% width. Ideal when you want to compare the original text and the translation at the same time — great for language learning.

Backend render styles

Separately from the display mode, each backend produces its translated output in a different way:

Backend	What the extension shows	Notes
Hosted Server (default)	Full image swap — the server inpaints over the original text and returns a fully translated image	Cleanest visual result; click-toggle restores the original image
OpenAI Vision (BYOK)	Canvas overlays drawn on top of the untouched art	Original pixels preserved; instant click-to-toggle
Ollama (local)	Canvas overlays drawn on top of the untouched art	Same as OpenAI — runs entirely offline