Test Gemini with Voice Input + Analysis + Structured Output

[amjadAwad95]

Goal: Evaluate Gemini’s voice/audio capabilities by sending an audio sample, analyzing the model output, and returning results in a structured format that matches our requirements.

---

Requirements

• Provide an audio/voice input to Gemini (recorded file or microphone stream).
• Convert audio to text (if needed) and/or use Gemini’s native audio input (depending on the API mode).
• Ask Gemini to analyze the voice input and return:
  • transcription
  • key information extraction (what we need)
  • summary
  • optional metrics (confidence, timestamps, language, etc.)
• Output must be structured (JSON) so it can be consumed by the backend.

---

Inputs

• Audio file (wav/mp3/m4a) OR live microphone recording
• Test prompt template (what we ask Gemini to do)
• Expected output schema (JSON fields)