STT

Voice is the most natural interface for hands-free scenarios — driving, cooking, customer support calls. STT bridges human speech and LLM text processing.

STT is a court stenographer — they listen to spoken words and type them out in real time for the record.

Scenario

User speaks into mobile app → streaming STT converts to text in 300ms → text sent to support agent → agent responds → TTS plays answer.

Solution

In Voice & Multimodal Agents, apply STT to this scenario: User speaks into mobile app → streaming STT converts to text in 300ms → text sent to support agent → agent responds → TTS plays answer. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to STT (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where STT happens in the code.

1from openai import OpenAI  # import dependencies2
3client = OpenAI()  # create API client4
5def transcribe(audio_file_path: str) -> str:  # define a reusable function6    with open(audio_file_path, "rb") as f:7        transcript = client.audio.transcriptions.create(  # call the API8            model="whisper-1",9            file=f,10            response_format="text",11        )12    return transcript  # return the result

• Treating STT as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for stt

Quick recap — the most important points from this module.