{
 "name": "Transcription timestamp accuracy on AMI meeting audio",
 "measured": "2026-09-24",
 "license": "CC BY 4.0",
 "source": "https://groups.inf.ed.ac.uk/ami/corpus/",
 "method": "AMI Meeting Corpus, meetings EN2002b and ES2004b, single distant microphone (SDM) mix, 69 minutes. Each file was sent unchanged to our production servers on 2026-09-24 with language set to English and diarization off. Reference: the corpus's human-annotated utterance start times; utterances opening with a filler (um, uh, mm) were excluded, leaving 1373.",
 "meetings": [
  {
   "id": "EN2002b",
   "minutes": 30,
   "eligibleOnsets": 844
  },
  {
   "id": "ES2004b",
   "minutes": 39,
   "eligibleOnsets": 529
  }
 ],
 "metric": "onset error in milliseconds: transcript start time minus human-annotated utterance start",
 "results": [
  {
   "engine": "Parakeet-TDT-0.6b-v3",
   "level": "word",
   "label": "Word-level",
   "eligible": 1373,
   "medianMs": 130,
   "p90Ms": 610,
   "biasMs": -40,
   "within100": 43.0,
   "within250": 77.6,
   "within500": 88.3,
   "matched": 742
  },
  {
   "engine": "Parakeet-TDT-0.6b-v3",
   "level": "segment",
   "label": "Segment-level",
   "eligible": 1373,
   "medianMs": 250,
   "p90Ms": 1430,
   "biasMs": -10,
   "within100": 24.3,
   "within250": 50.1,
   "within500": 64.3,
   "matched": 888
  },
  {
   "engine": "Whisper large-v3-turbo",
   "level": "segment",
   "label": "Segment-level",
   "eligible": 1373,
   "medianMs": 460,
   "p90Ms": 1490,
   "biasMs": 0,
   "within100": 20.8,
   "within250": 36.3,
   "within500": 51.9,
   "matched": 1084
  }
 ]
}