Writing · 2025
Comparing Speech-to-Text across Cloud, Edge and Local Deployments
Benchmarking Silero and Whisper against Azure and Google Cloud speech services, to find out where a low-latency drone command channel should actually run.
Sole-authored, Cal Poly Pomona · 2025
Real-time speech-to-text is the front door of a spoken command-and-control interface, and for a drone it has to be both fast and right — a late transcription and a wrong transcription fail in different but equally unacceptable ways.
This study evaluates two offline engines, Silero and OpenAI Whisper, across local, edge and local-cloud deployment paradigms, and compares them against hosted services from Azure (running Whisper) and Google Cloud. Measurements were taken across a range of hardware rather than a single reference machine, because the deployment decision is really a hardware decision.
The point of the comparison is that the three paradigms fail differently: cloud services buy accuracy with a network round trip and a hard dependency on connectivity, while offline engines trade some accuracy for latency you control. Which one is correct depends entirely on whether the link can be assumed.