Voice prompts walk operators through every inspection step. Captures asset ID, fluid levels, tire condition, and service status as DOT-ready structured slots, plus free-form defect notes. Runs entirely on the device.
The on-device voice guided inspection pipeline runs five Picovoice SDKs in a loop on the operator's device: Porcupine Wake Word, Rhino Speech-to-Intent, Cheetah Streaming Speech-to-Text, Orca Text-to-Speech, and Koala Noise Suppression. Porcupine listens for the wake word. Koala suppresses shop, ramp, and engine noise. Orca speaks each inspection prompt aloud. Rhino captures structured slots — asset ID, fluid levels, tire condition, service status — directly from speech. Cheetah transcribes free-form defect notes. Audio never leaves the device.
Shop bays, depots, and roadside walkarounds are loud. Koala Noise Suppression cleans air tools, idling engines, traffic, and generator noise before the audio reaches Porcupine, Rhino, and Cheetah. Koala suppresses background noise twice as effectively as RNNoise with the same compute footprint, which leaves headroom for the rest of the pipeline on embedded devices, legacy phones, and rugged handhelds from Honeywell or Zebra.
Porcupine Wake Word starts the inspection workflow when the operator says the chosen phrase. Drivers can train a branded wake word or always-listening commands, such as “Start inspection” or “Begin walkaround” in the Picovoice Console, and deploy them across mobile and embedded devices. Porcupine runs always-on at low CPU and battery cost so the rest of the pipeline only spins up when needed.
Rhino Speech-to-Intent captures structured DVIR slots — unit ID, oil condition, tire condition, service status — directly from speech. Most voice command systems run a two-step pipeline: speech-to-text produces a transcript, then a separate NLU model parses that transcript for intent. Each step accumulates error and compounds latency. Rhino infers intent and typed slot values directly from audio, holding higher accuracy even in noisy environments without hallucinations or compounding errors.
Cheetah Streaming Speech-to-Text transcribes free-form defect notes in real time. Per the open-source real-time transcription benchmark, Cheetah beats Google Cloud STT on word error rate and word emission latency across all tested languages, and outperforms Azure STT on several. It emits words at 590 ms median latency, typically one word behind the speaker, and requires less compute than any other local engine tested. Cheetah accepts custom vocabulary for fleet jargon and OEM part numbers, which raises accuracy on the words inspectors actually say.
Orca Streaming Text-to-Speech reads each inspection prompt aloud — “What's the oil level?”, “What's the tire condition?” — so the operator never has to look at the screen. Most high-quality TTS engines require hundreds of megabytes of RAM. Orca uses 29 MB peak memory, 10 to 50 times less than any natural-sounding on-device alternative, which leaves enough headroom to run all five engines on a single rugged tablet without OOM crashes.
A Driver Vehicle Inspection Report (DVIR) is a daily record that the U.S. Department of Transportation (DOT) requires for commercial motor vehicles. Voice guided inspection apps turn pre-trip and post-trip DVIR walkarounds into spoken workflows that capture assets, conditions, and status as structured slots, ready to push to Whip Around, Fleetio, Samsara, or Verizon Connect.
Heavy equipment on construction sites, in mines, quarries, and large farms needs daily inspection in places with no signal. The on-device pipeline runs on a commodity Android phone or rugged tablet from Honeywell or Zebra and captures structured condition data and free-form defect notes with no connectivity required. No proprietary inspection hardware required.
Plant inspectors, utilities crews, and HVAC technicians walk past boilers, pumps, motors, and panels that need a quick condition check. The on-device voice guided inspection pipeline prompts each asset, captures the reading and the inspector's free-form note, and works offline in basements, mechanical rooms, and remote substations where Wi-Fi doesn't reach.
Aviation pre-flight checks and marine vessel pre-departure inspections are read-aloud, respond-aloud workflows by nature. The voice guided pipeline matches that workflow and runs offline, critical at smaller airfields, offshore platforms, and harbors where connectivity is unreliable. The same structured slot capture handles airframe, engine, and safety checks.
A complete working recipe in Python. Open-source on GitHub. Runs 100% on-device.