test_speech_recognition_cross_origin.html

Enable keyboard shortcuts

<!DOCTYPE HTML>

<html>

<head>

  <meta charset="utf-8">

  <title>SpeechRecognition: a cross-origin track is fed as silence</title>

  <script src="/tests/SimpleTest/SimpleTest.js"></script>

  <script src="head.js"></script>

  <link rel="stylesheet" type="text/css" href="/tests/SimpleTest/test.css"/>

</head>

<body>

<div id="content"></div>

<pre id="test">

<script>

"use strict";

// The same speech as test_parakeet_e2e.html, loaded from another origin.

const SERVER_PORT = 8766;

const LANG = "en-US";

const CROSS_ORIGIN =

  "https://example.org/tests/dom/media/webspeech/recognition/test/";

const LISTEN_DURATION_MS = 15000;

SimpleTest.requestFlakyTimeout("waiting for parakeet inference to complete");

add_setup(async function setup() {

  await SpecialPowers.pushPrefEnv({

    set: [

      ["media.webspeech.recognition.enable", true],

      ["browser.ml.modelHubRootUrl", `http://localhost:${SERVER_PORT}/`],

      ["media.webspeech.recognition.model-download.prompt.testing", true],

      ["media.navigator.permission.disabled", true],

],

});

});

add_task(async function cross_origin_track_is_blanked() {

  ok(await ensureModelInstalled([LANG]), "Model installed successfully");

  const audio = await loadTestAudio(CROSS_ORIGIN + "kennedy-appolo.opus");

  const track = audio.mozCaptureStreamUntilEnded().getAudioTracks()[0];

  ok(track, "Got audio track");

  audio.currentTime = 0;

  await audio.play();

  const sr = new SpeechRecognition();

  sr.continuous = true;

  sr.lang = LANG;

  let speechstartFired = false;

  sr.onspeechstart = () => {

    speechstartFired = true;

};

  const transcript = await new Promise((resolve, reject) => {

    let text = "";

    sr.onresult = e => {

      for (let i = e.resultIndex; i < e.results.length; i++) {

        if (e.results[i].isFinal) {

          text += " " + e.results[i][0].transcript;

};

    sr.onerror = e => reject(new Error(e.error));

    sr.onend = () => resolve(text.trim());

    sr.start(track);

    setTimeout(() => sr.stop(), LISTEN_DURATION_MS);

});

  audio.pause();

  is(transcript, "", "A cross-origin track produces no transcript");

  ok(!speechstartFired, "No speech is detected in a cross-origin track");

});

</script>

</pre>

</body>

</html>