Observed arrival · 2026-10-11
WhistleLab puts a 16.9 MB speech model through its paces
WhistleLab compares the Whistle speech-to-text model with other speech models and offers browser-based word-error and device-fit tools.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Developers evaluating compact, on-device speech recognition
- Worth noticing
- Its device checker covers 17 build targets and labels the official measurement separately from its estimate.
Field notes
The page reports results across nine named datasets and 86,174 utterances, with separate entries for clean and accented LibriSpeech and two AMI conditions. It distinguishes model-file size from runtime memory, giving about 28 MB as the inference requirement. For latency, it identifies the 11.1 ms figure as the vendor's measurement on a 10-second clip and notes that the hardware was not specified.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue