Interview format (NormalizedInterview v2.0)

If you want to author interviews as JSON — for import via Upload Template or for source-controlled definitions — Ear3 uses a native format called NormalizedInterview v2.0. It’s the master representation from which the editor UI, the Pipecat runtime, and Qualtrics exports are all generated.

Minimal shape:

{
  "version": "2.0",
  "questions": [
    {
      "id": "QID1",
      "type": "voice_input_single",
      "text": "Tell me about your last onboarding.",
      "required": true,
      "order": 0,
      "next": "__continue__"
    }
  ],
  "metadata": {
    "preamble": "You are a friendly research interviewer…",
    "language": "en",
    "voiceId": "…"
  }
}

Question type values include voice_input_single, multiple_choice_single, likert_scale, and display_list. Multiple-choice choices can carry a transitionTo for conditional branching; text/voice inputs accept validation with minLength, maxLength, pattern.

References:

• Full spec — ear3.ai/docs/normalizedinterview-format-v20
• Open-source normalizer + Pipecat compiler — github.com/ear3-ai/voice-flow-format-normalize