--- title: Recognizing Text in Images framework: vision role: article role_heading: Article path: vision/recognizing-text-in-images --- # Recognizing Text in Images Add text-recognition features to your app using the Vision framework. ## Overview Overview One of Vision’s many powerful features is its ability to detect and recognize multilanguage text in images. You can use this functionality in your own apps to handle both real-time and offline use cases. In all cases, all of Vision’s processing happens on the user’s device to enhance performance and user privacy. Vision’s text-recognition capabilities operate using one of these paths: For example code using the fast path, see Extracting phone numbers from text in images. For example code using the accurate path, see Structuring recognized text on a document. Using either path, you may optionally apply a language-correction phase based on Natural Language Processing (NLP) to minimize the potential for misreadings. note: Using Vision’s text-recognition features is similar to performing other Vision operations, where you perform computer vision requests on an image and retrieve the resulting observations. If you’re new to the Vision framework, see Detecting Objects in Still Images. Perform a Text-Recognition Request Vision provides its text-recognition capabilities through VNRecognizeTextRequest, an image-based request type that finds and extracts text in images. The following example shows how to use VNImageRequestHandler to perform a VNRecognizeTextRequest for recognizing text in the specified CGImage. // Get the CGImage on which to perform requests. guard let cgImage = UIImage(named: "snapshot")?.cgImage else { return } // Create a new image-request handler. let requestHandler = VNImageRequestHandler(cgImage: cgImage) // Create a new request to recognize text. let request = VNRecognizeTextRequest(completionHandler: recognizeTextHandler) do { // Perform the text-recognition request. try requestHandler.perform([request]) } catch { print("Unable to perform the requests: \(error).") } note: VNRecognizeTextRequest uses the accurate path by default. To select the fast path, set the request’s recognitionLevel property to VNRequestTextRecognitionLevel.fast. Process the Results After the request handler processes the request, it calls the request’s completion closure, passing it the request and any errors that occurred. Retrieve the observations by querying the request object for its results, which it returns as an array of VNRecognizedTextObservation objects. Each observation provides the recognized text string, along with a confidence score that indicates the confidence in the accuracy of the recognition. func recognizeTextHandler(request: VNRequest, error: Error?) { guard let observations = request.results as? [VNRecognizedTextObservation] else { return } let recognizedStrings = observations.compactMap { observation in // Return the string of the top VNRecognizedText instance. return observation.topCandidates(1).first?.string } // Process the recognized strings. processResults(recognizedStrings) } If you’d like to render the bounding rectangles around recognized text in your user interface, you can also retrieve that information from the observation. The rectangles it provides are in normalized coordinates. To render them correctly in your user interface, convert CGRect instances from normalized coordinates to image coordinates by using the VNImageRectForNormalizedRect(_:_:_:) function as shown below. let boundingRects: [CGRect] = observations.compactMap { observation in // Find the top observation. guard let candidate = observation.topCandidates(1).first else { return .zero } // Find the bounding-box observation for the string range. let stringRange = candidate.string.startIndex..