The text translate page is a separate entry from the camera page for cases where the source is already text. Open it from the bottom of the camera page and paste, type or dictate a paragraph; the same ML Kit engine returns a translation.
Why a separate page
Camera OCR is optimised for images. When the source is already text — an email snippet, a chat message, a pasted paragraph — the OCR step is overhead. The text translate page skips OCR entirely and feeds the input directly to the ML Kit translator.
It also lets you batch-translate by copying and pasting a longer paragraph rather than photographing it. For dense text with no convenient page to photograph, this is faster and more accurate.
How it works
- Shared language state. The text page reads and writes the same source/target language pair as the camera page. Selecting a target language in one place applies to the other; switching back and forth does not require re-selecting languages.
- Auto-detect for source. If the source is set to "Auto Detect", the app runs a Dart-layer character-set heuristic on the input text to choose between ML Kit-supported languages. Pure Latin defaults to English; Han characters default to Simplified Chinese; mixed scripts follow the majority character class.
- Single-pass translation. The whole input is sent to ML Kit in one call. The result replaces the input area below the button; clear the input to translate again.
- Keyboard handling. Tapping Translate dismisses the on-screen keyboard so the result region is not hidden behind it.
When to use it
- Long paragraphs. Type or paste multi-sentence text. Photographing a screen full of text introduces OCR noise; pasting the text does not.
- Chat snippets. A translate a short message from a friend without leaving the conversation to open the camera.
- Screens with text. Pull text from a screenshot via system share and paste it into the text translate page for a cleaner translation than OCR.
- Voice dictation. Use the system dictation key to to dictate text instead of typing.
Limits
- No formatting preservation. Plain text only. Markdown, bold, italic and other formatting are dropped.
- No history auto-save. Unlike camera captures, text translations are not auto-saved. The result stays in the result region until you tap Clear or navigate away.