ATIS-style and AWOS-style spoken loops · on-frequency use needs authority approval
KBM AeroVox reads the airport's live METAR, composes the broadcast script by fixed phraseology rules, and voices it in English, Spanish, French or Portuguese. It is re-voiced on the ATIS cycle when the weather changes. The voice is synthetic; the words are not generated.
The AI is the voice, never the author.
Interactive demo
Why AeroVox
AeroVox slots in next to the broadcast you already run. Same observation data, same ATIS cycle, same phraseology rules, delivered in a clear, natural voice.
When an airport is explicitly configured for live operation, broadcasts are polled every half hour by default and re-baked when the configured weather signature changes. Immediate special-observation handling requires an authoritative event feed.
Every broadcast is generated by the same deterministic rule set: same phraseology, same unit conventions, same cadence. The AI is the voice, never the author; with the required production audit controls, the output can be audited line-by-line.
The same observation is spoken in English, Spanish, French, and Portuguese today, with native phrasing and a native voice. Serve international, military exchange, and flight-school traffic from one feed.
AeroVox reads the same METAR the official broadcast already uses and follows your facility's phraseology. No workflow change for controllers, FBOs, or airport ops: it is the broadcast you have, in a clearer voice.
Architecture
METAR pulled from FAA / NWS at the source. No human re-keying, no scraping intermediaries.
Raw observation decomposed into typed fields (wind, visibility, ceiling, altimeter, remarks) with units normalized.
Fixed templates in standard ATIS/AWOS phraseology compose the script. No model invention, no paraphrase: every clause maps to a field.
Same structured fields rendered into English, Spanish, French, and Portuguese against locale-aware phraseology tables.
Neural voice reads the approved script at measured ATIS pace. Re-baked on the ATIS cycle (default 30 min).
Every spoken weather value traces to a field in the authoritative observation; runway, approach and the information letter come from airport configuration. The AI is the voice, never the author.
How it works
Pulls the airport's real, current METAR: the same data the official broadcast uses.
Converts to proper ATIS phraseology by fixed rules, in each supported language. The AI is the voice, never the author.
Speaks it in a clear voice at a measured ATIS pace, updated on the cycle like a real reissue.
Deployment options
Same deterministic engine, same voice quality, same multilingual coverage. Three operating models for three buyer realities.
Managed AeroVox running in KBM infrastructure. We handle the METAR ingestion, the deterministic phraseology engine, the voice synthesis cache, and the broadcast endpoint.
AeroVox installed in your data center or private cloud. Your METAR feed, your network boundary, your retention policy. We deliver the container stack and the operations runbook.
Fully isolated install with no outbound network. Voice models, phraseology rules, and METAR ingestion all run inside the enclave. Designed for federal, military, and sovereign-AI requirements.
ROI calculator
Adjust to your situation. All inputs are estimates. This is not a quote.
First-deployment program
We are taking a small number of first-deployment partners. You get the engineering attention of a pre-product-market-fit team; we get the operational data to prove the case.
Pilot fee covers integration and the 90‑day evaluation. Fully credited against year-one contract if you convert. Walk-away at day 90 with no obligation. Written terms before any signature.
The communications gap
The data is live. The weather sensors are live. The voice is decades old.
Most other layers of the aviation information stack have been digitized and automated.
METAR data flows in real time from sensors to cockpits. Aircraft know their exact position.
Passengers check gate status on their phones.
But the broadcast a pilot copies (the voice that gives wind, visibility, ceiling, altimeter)
is still produced by a system designed in the 1970s, voiced by a machine that sounds like it.
At many airports, that broadcast is re-recorded only when someone remembers to do it.
AeroVox is a first step toward closing that gap: a modern aviation communications layer
that treats the spoken broadcast as data, not a recording.
Platform roadmap
ATIS-style weather is the starting point, not the ceiling. AeroVox is being built as an airport communications platform: the same deterministic engine and the same multilingual voice layer, applied to every spoken broadcast a field generates. Voice is the feature. Operational automation is the business.
Live METAR, deterministic FAA / ICAO phraseology, native voice in four languages, broadcast on the ATIS cycle. — airports on real weather in the demo above.
Read field NOTAMs aloud on demand, in any of the supported languages. Same deterministic phraseology approach: the model voices the authoritative NOTAM text, never paraphrases it.
Spoken RCR / runway contamination / braking action reports built from the existing surface observation, in the same voice and cadence as the weather broadcast.
Bird activity, equipment outages, special-event advisories, voiced from the operations log against a controlled alert vocabulary, not free-form generation.
Non-METAR field information (fuel availability, FBO status, customs hours, after-hours contact) in the same multilingual voice layer as ATIS.
Terminal-side announcements in multiple languages from the same engine that voices the airside broadcast. One source of truth, airside and landside, one voice identity.
Part of the KBM Nexus operational-intelligence family. Roadmap items above are not shipping today · no completed pilots · sequencing follows first-customer demand.
Why now
The infrastructure is ready, the pain is documented, and the pressure on airports is increasing. Here's what's changed.
Natural-sounding synthetic voice became practical at scale around 2022–2023. Broadcast-quality speech that once required a recording studio can now be generated on demand on commodity cloud infrastructure, in multiple languages.
The technology is now practical.The post-pandemic aviation labor market is tight across every function. Small and regional airports are operating leaner than at any point in the last decade. Manual broadcast re-recording, an after-thought task in well-staffed operations, is now genuinely burdensome.
GA airport staffing pressure has intensified since 2019.A substantial share of U.S. commercial passenger volume involves non-native English speakers, and that share is rising at regional airports with growing international charter, military exchange, and flight-school populations. English-only AWOS was adequate when international traffic was rare.
A rising share of U.S. passengers are non-native English speakers.FAA's NextGen program, the DOT's digital infrastructure agenda, and state aviation modernization funds are directing capital toward this problem. Defense innovation programs (AFWERX, DIU) are actively evaluating AI-augmented voice systems for airfield operations.
FAA NextGen + state modernization funds actively deployed.The question airports will face in the next 24 months is how to modernize the spoken broadcast. AeroVox is built to be one answer to that question.
Common questions
Short, direct answers. For anything specific to your facility, request a briefing.
Something we didn't cover? Send us the question directly.