A floating button sits on top of any app. Tap it to translate the text you can see, or ask an AI a question about what's on the screen. That's the core of it.

TRANSLATE SCREEN

The translator reads text from whatever's currently displayed (English, Vietnamese, Chinese, Japanese, Korean, Hindi, plus 70+ Latin-script languages) and shows the translation in place, without copying or switching apps.

You pick the engine per job:
- ML Kit, offline. No internet, no quota. First time you use a language, it downloads once, then runs locally forever.
- Google Translate. The default for everyday text.
- FTouch Translate, an in-house model tuned.
- ChatGPT Fast or ChatGPT Pro, when you want phrasing that sounds like a person wrote it.
- Claude Haiku or Claude Sonnet 4.6, for long passages, idioms, and prose where nuance matters.
- Gemini Flash or Gemini 2.5 Pro, the strongest option for CJK and low-resource languages.

Two ways to read the result. Overlay mode stamps the translation directly over the original block, so menus and chats stay where they were. Panel mode opens a scrollable side window for articles and longer reads. Switch between them mid-session.

Region Translate lets you draw a rectangle, or freehand lasso, around just the speech bubble, sign, or paragraph you actually want. Draw several regions in one pass and confirm them together. A persistent pill shows the current source and target language, and gets out of the way when you're done.

Manga Mode swaps in a cloud OCR built for vertical Japanese text and irregular speech bubbles. Game Mode makes the overlay touch-transparent so taps fall through to the game underneath. Auto-translate loops every 1 to 10 seconds when you want subtitle-style continuous translation for a video. Smart-merge collapses fragmented OCR boxes back into proper paragraphs; the noise filter strips timestamps and watermarks before sending text to the engine.

ASK SCREEN

Ask Screen takes a capture of what you're looking at and sends it, with your question, to an AI model. The model can read the text, identify objects in an image, summarize a long page, explain a chart, or translate with context that the plain translator can't see.

Three models, picked from a sheet when you start a chat:
- Gemini, for quick everyday questions.
- Claude Haiku, when you want a more careful reading.
- GPT-4.1, for harder reasoning.

Conversations stay open. You can keep asking follow-up questions about the same screenshot, or send a fresh capture mid-thread. History persists across sessions, and a per-app context hint tells the model which app you were just in, so its answers stay relevant.

EVERYTHING ELSE

The same floating button hides a few smaller tools you may or may not use:

Night Filter has four warmth levels, five color presets including a custom HSV picker, an adaptive mode that follows your ambient light sensor, and a per-app blacklist so the filter pauses on apps you choose.

A magnifier with adjustable zoom and preview size, switchable between realtime refresh (smooth) and periodic refresh (kinder to battery).

A screenshot editor that captures, then lets you drop a gradient background, device frame, annotations, watermark, padding, or shadow on it before exporting as PNG, JPG, or WebP.

Sticky notes that stay pinned on screen. A keep-awake toggle. A touch blocker with a corner-unlock gesture. Three gestures on the floating button (tap, double-tap, long-press), each rebindable to any of the main features above.

UI in English, French, Spanish, German, Portuguese, Russian, Hindi, Thai, Korean, Chinese, Japanese, Vietnamese, and a few more. The translator also remembers your last source/target pair per app.

PRIVACY

Screen captures are sent only when you trigger a feature that asks for them, and only to the engine you picked for that feature. The offline ML Kit option keeps everything on the device. There are no third-party trackers.