Skip to content
Testzilla

Test Web Voice

Web Voice tests a browser voice agent over a real WebRTC session — the same path a visitor takes when they click a voice button on your site. Testzilla launches a headless Chromium instance, loads the provider’s web-call SDK, establishes a WebRTC call with a just-in-time access token, and runs the conversation with synthesized tester audio. The full pipeline is exercised: WebRTC media negotiation, the agent’s ASR, LLM, and TTS, in a browser context. This validates the web embed, distinct from a phone call.

Before you start

Section titled “Before you start”
  • A Testzilla account (sign up).
  • An agent deployed as a browser voice widget on Retell, Vapi, or ElevenLabs.
  • The agent ID for your provider.

Set it up

Section titled “Set it up”

  1. Create a project, or describe the whole thing to Tessie in plain language.

  2. Add a Web Voice channel

    Click + Channel, pick the Web Voice adapter, and select your provider:

    • Retell — supply the Voice Agent ID; the web-call token is minted server-side.
    • Vapi — supply the assistant details; the browser SDK connects directly to the agent.
    • ElevenLabs — supply the ElevenLabs agent ID (and a conversation token for private agents).

    Optionally set the sample rate (default 24000 Hz) and the tester’s TTS voice.

  3. Write a test with the scenario and pass/fail criteria.

  4. Run it and watch it on the Queue page.

  5. Read the result — transcript (reconciled from the provider’s post-call transcript), verdict, score, analysis, and collected data.

Next steps

Section titled “Next steps”