Help center

English-speaking adults (18+) with non-standard speech who have access to an Android mobile device or tablet (Android 8+). Project Relate is currently available to users in AU, CA, GH, IN, NZ, UK, US.

Yes. You can create Custom Cards so the app understands any proper nouns or personal vocabulary. If you already have a personalized Project Relate model, it will require 24 hours for the Custom Card recordings to be added to your model.

For best success, record the same proper noun across 3-5 cards.  For example, if ‘Gloucester’ is the proper noun, record three cards, like:

“My mom lives in Gloucester.” 

“Gloucester has a lot of fishing boats.” 

“I went to the Gloucester waterfront festival.”

Each word is paired to the audio and used to train the personalized speech model, so it’s important to say each word as it is written. That being said, you can skip as many cards as you need to. Use the Custom Cards feature to record any phrases in your own words.

Yes. Whether you're looking to use Project Relate for yourself or you have a client who could benefit from the app, you can find more information in our user guide and tutorial video.

Project Relate is intended to be a communication tool in combination with active listening, gesture tracking, and asking for clarification. Not every personalized model will have the same level of helpfulness. After the necessary phrases are recorded, Project Relate will provide users with a personalized speech model with a recognition score of Low, Medium, or High. Additional recordings can potentially help improve a personalized model. 

Before you take the time to record 500 phrases, here are a few things you should know. If your voice is changing quickly, Project Relate will become less useful over time. Also, if your voice is severely impacted (even my closest friends and family don't understand me), Project Relate may not be able to understand you with high accuracy.

Some functions of the app require WiFi or data access, but some do not. When you’re first training the app to understand your voice, you can record without WiFi. Recordings will be saved on your device and uploaded when WiFi is available, or you’re able to connect to mobile data.

Once you have your personalized speech model, you can access certain features with or without WiFi. 

• The Listen feature does not require an internet connection

• The Repeat function requires an Internet connection.

• The Assistant feature requires an Internet connection.

Listen: To transcribe speech during a conversation. 

Repeat: Short phrases are repeated via sound-by-sound interpretation of organic speech into synthetic speech.

Assistant: When using Google Assistant to get answers, control your phone, and more.

Good quality recordings are critical for teaching Relate to understand you. Here are some tips for creating the best recordings:

• Place the phone 6-18 inches from your mouth, with either the screen or bottom of the phone pointed towards you. Try to keep the phone closer to you if you have a quiet voice.

• Do not cover the top or bottom edge of the phone while holding it as this may block the microphone.

• Avoid background noise like televisions or loud fans. Make sure no one else is speaking around you at the time you are recording.

• Try to avoid recording in rooms with hard surfaces that echo.

After you record 500 phrases, a personalized speech model will be created for you. This can take up to 24 hours once these phrases have been uploaded. You will need Wifi or a data plan for these phrases to be uploaded. You can expect a notification on your device and another within the app to alert that your speech model is ready. You will also receive a “Speech recognition score” on this notification, which will look something like this:

People who use Relate can choose whether or not to use their speech therapy strategies. The important thing is to be consistent. If a person uses these strategies when recording Relate phrases, they should continue to use the same strategies while using the app.

The Relate Keyboard is a keyboard that you can enable to allow you to dictate in Google Docs, Email, Google Translate, or any other location in which you would use a keyboard, all using your voice.

• Does the Relate Keyboard require an internet connection? The Relate Keyboard feature does not require an internet connection.

• How do I access the Relate Keyboard?

1. Go to your Android device's Settings > System > Languages & input > On-screen keyboard (or Virtual keyboard on Android 10).

2. Select Manage on-screen keyboards and enable Relate.

3. Open the app where you want to type. For example, Google Docs.

4. Tap an area where you can enter text. The on-screen keyboard should appear.

5. Tap the small keyboard icon Key in the bottom right corner to bring up the input selection menu, then tap Relate.

6. Tap the keyboard icon again to switch back to the normal keyboard.

Note that these steps are for Pixel users and may vary somewhat depending on mobile device.

Project Relate

You record at least 500 phrases in the Project Relate app and receive a custom speech recognition model that you can use to speak to your Android device.

Project Relate is available only on Android mobile devices. 

In addition to a preset list of phrases to record, you can also create Custom Cards to record phrases in your own choice of words. These count toward your 500 recordings. You can create and record as many Custom Cards as you’d like.

Recordings are used to create your personalized speech model and also to benefit speech recognition research by Google.

Project Euphonia

Project Euphonia is the research initiative that preceded the development of our Android app, Project Relate.  Learn more about Project Euphonia at g.co/Euphonia.

You record 300 phrases for Project Euphonia and receive $60 (or local currency equivalent) in the form of an e-gift card.

Recording is supported on any desktop, computer, tablet, or mobile device with Chrome or Safari. 

Euphonia uses a web tool called ChitChat with a preset list of phrases to record. The web tool can be accessed via desktop or mobile device. 

All recordings are used for the sole purpose of benefiting speech recognition research by Google.