Speaker identification

The Speaker Identification API is used to determine who is speaking for a given media file. It is intended to be used to identify speakers who have been previously enrolled with the AI API. The Speaker Identification API segments an audio clip into a sequence of utterances, each corresponding to a unique speaker. It then attempts to determine the identity of each speaker based upon their voice print. In cases where the speaker is ambiguous or unknown, utterances are marked with a status of UserNotIdentified.

Another use case for using Speaker Identification API is with multi-stream audio files. Each individual on a multi-stream audio file is a different speaker, but you don't know which stream belongs to which speaker. In this case, you can send each individual stream to the Speaker Identification API, and let the API identify the speaker for you using the audio stream.

Synchronous vs asynchronous speaker identification

Identifying a speaker only takes the first few minutes (approximately 1-2 minutes) of audio. If you wish to identify the speaker, you can simply cut your audio down to 2 minutes and submit that to the synchronous API:

GET /ai/audio/v1/speaker-identify

The asynchronous API, however, will not only identify the speaker, it will also utilize speaker diarization as well. This means you can use the asynchronous API to not only identify the speaker, but also see what they said. This output is similar to the diarization API with the added benefit of identifying the speaker. Since the entire media file is being analyzed, this API uses an asynchronous operation.

GET /ai/audio/v1/async/speaker-identify

Identifying speakers

Request parameters

To identify speakers using the Speaker Identification API, one must formulate a request using the following request parameters:

To properly identify speakers in a media file, speakers must have been previously enrolled so that their voice print can be compared to the speakers in the audio file. It also relies on the developer having some knowledge of who the likely speakers are in the media file being processed, and providing a list of those speakers in their request.

Sample code

After you have setup a simple web server to process the response, copy and paste the code from below in index.js and makesure to edit the variables in ALL CAPS to ensure your code runs properly.

const RC = require('@ringcentral/sdk').SDK;
require('dotenv').config();

MEDIA_URL   = process.env.RC_MEDIA_URL;
WEBHOOK_URL = '<INSERT YOUR WEBHOOK URL>';
SPEAKERS = [
    "ringcentral_test",
    "speakerId2"
]

// Initialize the RingCentral SDK and Platform
const rcsdk = new RC({
    'server':       process.env.RC_SERVER_URL,
    'clientId':     process.env.RC_CLIENT_ID,
    'clientSecret': process.env.RC_CLIENT_SECRET
});

const platform = rcsdk.platform();

// Authenticate with RingCentral Developer Platdorm using Developer's JWT Credential
platform.login({
    'jwt': process.env.RC_JWT
});

// Call the Speaker Enrollment API right after login asynchronously
platform.on(platform.events.loginSuccess, () => {
    identifySpeakers();
})

async function identifySpeakers() {
    try {
        console.log("Enrolling speaker using RingCentral Enrollment API");
        let resp = await platform.post("/ai/audio/v1/enrollments?webhook=" + WEBHOOK_URL, {
            "contentUri":   MEDIA_URL,
            "encoding":     "Mpeg",
            "languageCode": "en-US",
            "source":       "RingCentral",
            "audioType":    "Meeting",
            "speakerIds":   SPEAKERS
        });
        console.log("Job is " + resp.statusText + " with HTTP status code " + resp.status);
    } 
    catch (e) {
        console.log("An error occurred : " + e.message);
    }
}

Example response

{
    "utterances": [
        {
            "start": 0,
            "end": 6,
            "speakerId": "ringcentral_test",
            "confidence": 35.42294
        },
        {
            "start": 6,
            "end": 12,
            "speakerId": "ringcentral_test",
            "confidence": 36.98796
        },
        {
            "start": 12,
            "end": 18.0,
            "speakerId": "ringcentral_test",
            "confidence": 25.51731
        }
    ]
}

Each utterance contains the following information and parameters: