---
title: "Can Nova-3 understand air traffic control?"
description: "A live stress-test of Deepgram Nova-3 on real ATC tower audio: baseline vs keyterm prompting, side by side, with word-level confidence."
url: "https://davidezambiasi.dev/projects/atc-live-transcript"
---

typescript · deepgram · next.js

# Can Nova-3 understand air traffic control?

A live stress-test of Deepgram Nova-3 on real ATC tower audio: baseline vs keyterm prompting, side by side, with word-level confidence.

Published July 21, 2026 · 1 min read

[Repository ↗](https://github.com/soos3d/atc-live-transcript)

## The problem

ATC radio is close to a worst case for speech-to-text: 8 kHz AM audio, squelch noise, clipped phraseology at speed, and callsigns no general model has seen enough of. That makes it a sharp benchmark for what Deepgram’s keyterm prompting actually buys you.

## What I built

A pipeline that takes a LiveATC stream, converts it with ffmpeg, and feeds identical audio bytes to two Deepgram Nova-3 websockets — one baseline, one with keyterm prompting. A Next.js UI shows both live transcripts with word-level confidence coloring and diff highlights, plus an audio relay of exactly what the model hears, so you can verify by ear.

## Why the judgments hold

I hold a commercial pilot certificate and CFI, so the keyterm lists and the accuracy calls are grounded in how these radio calls are actually spoken — the demo measures a real domain, not a synthetic benchmark.

DZ

Davide Zambiasi

Senior Developer Relations Engineer

I ship the docs, guides, and example apps that get developers building, and I hire for the roles that do the same.

[GitHub](https://github.com/soos3d)[LinkedIn](https://www.linkedin.com/in/davide-zambiasi)[Blog](https://blog.davideai.dev)

[

Preflight Agent: durable AI flight briefings

Previous Page

](/projects/preflight-agent-restate)[

Migrating a production bot to Nebius Token Factory

Next Page

](/projects/translator-bot-token-factory)