The following demo scenes can demonstrate how integrating Composer into your Voice SDK-enhance scene can enhance the way your users interact with your game.

To run these demo scenes, you’ll need to set up a composer graph like the following:

The Graph

Each composer graph is created based on the app you’re working on. However, for these demo scenes, the composer graph will be created based on Tutorial 2: Adding Voice Experiences to Your App with Custom NLP. This sets up the intents and entities used by these demos.

Note: You can also import the demo app from the composer_colours_demo_export.zip file in the /Oculus/Voice/Features/Composer/Samples/ folder in your Unity Assets folder.

Creating a Composer Graph

Create a new Wit app as shown in the tutorial.

In Wit, select Composer [BETA] on the left, or click the Composer icon to open it.

Create the graph. For more information on creating graphs, see the Wit.ai Composer instructions⁠. The graph itself begins with a check for an existing update to the context map. If there is one, it outputs the value of updated_info. This is used in the Context Map demo. Otherwise, it checks for voice input containing a set_color intent.

The flow of the graph is as follows:

A Input module: introduces the conversation flow by collecting an input utterance and moves the user to the B decision module.

B Decision module: checks for the context map value of updated_info, goes to C if present, otherwise it goes to D.

C Response module: responds with the value of updated_info using the text Context map now has “{updated_info}.” This path dead ends here.

D Decision module: checks the intent of the input utterance. If it has the color_set intent with ≥0.8 confidence, it goes to E. Otherwise it goes to H.

E Context module: Saves the entity color to the context map and goes to F.

F Decision module: checks for a color entity. If present, it goes to G. Otherwise, it goes to H.

G Response module: A valid color should be present here. The module says As you can see, I have turned {color[0].value}, using the value of the first entry of the color entity. This path ends here.

H Response module: no color has been provided or detected. It responds with Well you have to tell me a color to change it to! This path ends here (although you could change the graph to go back to the beginning).

Once you’ve completed this, open the SingleCharacterDemo/SingleCharacterDemo - WitConfiguration.asset inside the /Assets/Oculus/Voice/Features/Composer/Samples folder and copy the Server Access token from your Wit.ai app to that configuration file.

Demo 1: Single Character

In this demo, you can press the Activate button and speak a phrase such as “make yourself red”. It will be processed through composer using the flow A > B > D > E > F > G > H

If no phrase is recognized, it will instead terminate at G.

Demo 2: Double Characters

This scene demonstrates using two composer services. Activate the left or right to focus on those shapes. This demo otherwise operates the same as the Single Character demo.

Demo 3: Mixed Processing

This scene demonstrates the difference between using Composer and simply parsing the intents and entities as is done in the non-Composer shapes tutorial, simply by toggling between the two options. Using Composer automatically triggers the Text-to-Speech component, while the non-Composer iteration simply displays text.

Demo 4: Context Map

This scene demonstrates modifying the context_map to pass data to Composer.

When you select Update Context Map, it populates the context_map. The current context_map is displayed at the top right of the screen. Select Activate to speak an utterance and send the utterance (and the updated context_map) to Composer. It will follow the flow of A > B > C and will not pay attention to your utterance, as long as updated_info is in the context_map.