I’ve run into this exact problem before, so here are some practical solutions to get past the "Waiting for the text recognition model to be downloaded" error:

1. Ensure Google Play Services is Up-to-Date #

ML Kit depends on Google Play Services to handle model downloads. First, check if the device has the latest version installed. You can add a quick check in your app to prompt users to update if needed:

val playServicesStatus = GoogleApiAvailability.getInstance().isGooglePlayServicesAvailable(requireContext())
if (playServicesStatus != ConnectionResult.SUCCESS) {
    GoogleApiAvailability.getInstance().getErrorDialog(requireActivity(), playServicesStatus, 9000)?.show()
}

2. Manually Trigger Model Download #

The auto-download via the manifest meta-data tag doesn’t always work reliably. Try explicitly requesting the model download and listening for the result:

val textRecognizer = TextRecognition.getClient()
textRecognizer.downloadModelIfNeeded()
    .addOnSuccessListener {
        // Model is ready—start your text recognition here
        performTextRecognition()
    }
    .addOnFailureListener { exception ->
        Log.e("MLKitError", "Model download failed: ${exception.message}", exception)
        // Handle the failure (e.g., show a retry button to users)
    }

3. Double-Check Your Dependencies #

Make sure you have the correct ML Kit dependencies in your module-level build.gradle file. Using outdated versions can cause unexpected issues:

// Use the latest stable version of ML Kit Text Recognition
implementation 'com.google.firebase:firebase-ml-vision-text-recognition:24.1.0'

Also, confirm that you’ve applied the Google Services plugin in your project-level build.gradle:

plugins {
    id 'com.android.application'
    id 'com.google.gms.google-services' // Required for Firebase/ML Kit integration
}

4. Verify Network and Storage #

• The device needs an active internet connection to download the model (even with the pre-download tag, sometimes it requires a post-install connection to trigger the download).
• Ensure there’s enough free storage space—on-device text recognition models are roughly 200-300MB, so low storage can block the download.

5. Test on a Physical Device #

Emulators often have issues with Google Play Services syncing and model downloads. If you’re testing on an emulator, switch to a real Android device with Google Play Services installed.

6. Check Region Availability #

Some ML Kit models are restricted in certain regions. Make sure your device’s system region is supported for the on-device text recognition model.

Original Question #

I'm integrating Google Machine Learning Kit's On-device text recognition feature into my Android app. I followed the official guide to implement it, but every time I try to detect text, it throws the exception:
"Waiting for the text recognition model to be downloaded. Please wait."

Waiting doesn't help, and retrying after 10 seconds still results in the same error.

I've added the following meta-data tag inside the application tag in my Manifest:

<meta-data android:name="com.google.firebase.ml.vision.DEPENDENCIES" android:value="text" />

This tag is supposed to pre-download the recognition model when the app is installed.

Additional note: I tried the sample app (text-recognition/final) and it has exactly the same issue.

Am I missing a configuration item, or is there an API bug? Any help would be greatly appreciated.

内容的提问来源于stack exchange，提问作者M.Paunov