EME Workflow

What EME is

EME (Encrypted Media Extensions) is the W3C JavaScript API browsers expose for DRM-protected playback. It's deliberately key-system-agnostic: the same code drives Widevine, PlayReady, or FairPlay depending on what the browser supports.

EME deliberately offloads two things to the application:

1. Network calls between the CDM and the licence server. The browser hands you opaque message bytes; you POST them; you feed the response back.
2. Init data extraction, when needed. For CENC content, the browser extracts PSSH and gives you the init data for free.

This split is why EME integrations have so many places to go wrong — every byte that crosses the application is a place where wiring can break.

The complete flow

Code skeleton

The shape of the integration in real JavaScript:

async function play(video, src, licenseUrl) {
  // 1. Capability check
  const access = await navigator.requestMediaKeySystemAccess(
    "com.widevine.alpha",
    [{
      initDataTypes: ["cenc"],
      audioCapabilities: [{ contentType: 'audio/mp4; codecs="mp4a.40.2"' }],
      videoCapabilities: [{
        contentType: 'video/mp4; codecs="avc1.640028"',
        robustness: "SW_SECURE_CRYPTO", // L3-style; "HW_SECURE_ALL" for L1
      }],
    }]
  );

  // 2. MediaKeys + bind
  const mediaKeys = await access.createMediaKeys();
  await video.setMediaKeys(mediaKeys);

  // 3. Source
  video.src = src;

  // 4. Encrypted event drives session creation
  video.addEventListener("encrypted", async (e) => {
    const session = mediaKeys.createSession("temporary");
    session.addEventListener("message", async (event) => {
      const resp = await fetch(licenseUrl, {
        method: "POST",
        headers: { "Authorization": `Bearer ${appToken}` },
        body: event.message,
      });
      if (!resp.ok) throw new Error("licence failed");
      const buf = await resp.arrayBuffer();
      // 7. Feed back to CDM
      await session.update(buf);
    });
    // 4. Build the request
    await session.generateRequest(e.initDataType, e.initData);
  });

  await video.play();
}

Session types

EME defines three session types, set when you call createSession(type):

Robustness strings

The robustness field in videoCapabilities is how the application asks for a security level on Widevine:

• "" (empty) — let the CDM decide.
• "SW_SECURE_CRYPTO" — L3-equivalent: software CDM with software crypto.
• "SW_SECURE_DECODE" — software CDM, decode also software, but secure path.
• "HW_SECURE_CRYPTO" — keys in TEE; decode in rich OS (≈ L2).
• "HW_SECURE_DECODE" — keys + decode in TEE for video (≈ L1 video path).
• "HW_SECURE_ALL" — full L1 — keys, decode, all in TEE.

UHD policies typically demand HW_SECURE_ALL. If the device can't satisfy the requested robustness, requestMediaKeySystemAccess rejects.

Common error sites