text_sight library
Live, on-device text recognition — Apple Vision on iOS, ML Kit on Android.
Import only this file. The live entry points are TextSightView with a
TextSightController; the static one-shot is TextSight. Either way
recognition yields TextSightCaptures of RecognizedLines. TextSightModel
reports and controls when the on-device model is ready.
Classes
- ModelDownloading
- The unbundled ML Kit model is being fetched from Google Play Services.
- ModelReady
- The recognition model is present; recognition will produce results.
- The model cannot be made ready, so recognition will not produce results.
- RecognizedElement
- A sub-line recognition unit (typically a word) with its own text and box.
- RecognizedLine
- A single recognized line of text with its location and confidence.
- TextSight
-
One-shot, still-image text recognition — the static counterpart to the live
TextSightController. - TextSightCapture
- The result of one recognition pass — every RecognizedLine found plus the size of the image they were located in.
- TextSightController
- Configures and drives a live camera recognition session.
- TextSightModel
- Controls and observes readiness of the on-device recognition model.
- TextSightOptions
- The source-agnostic recognizer configuration shared by both drivers.
- TextSightReadinessState
-
The readiness of the on-device recognition model, reported by
TextSightModel. - TextSightView
- A live camera preview that recognizes text, driven by a TextSightController.
Enums
- Why the recognition model could not be made ready (see ModelUnavailable).
- RecognitionLevel
- How aggressively text is recognized, trading latency for accuracy.
Typedefs
- TextSightOverlayBuilder = Widget Function(BuildContext context, TextSightCapture capture, BoxConstraints constraints)
-
Builds an overlay painted over the live preview from the latest
capture.