Voice AI BenchmarksBy

ELEVENLABSSPEECH-TO-TEXT146,620 SAMPLES / 30 DAYSLAST RUN AUG 28, 2026, 21:30 UTC

official resource

Scribe v2 Realtime speech-to-text benchmarks

Scribe v2 Realtime is ElevenLabs' streaming speech recognition model.

Time to Final Segment#7 / 24
120ms
Word Error Rate#15 / 28
5.5%
Time to First Token#22 / 24
2157ms

Overview

ElevenLabs provides Scribe through a real-time transcription API with multilingual, voice-activity and keyterm support.

Scribe v2 Realtime is tested every day on fixed public audio — clean, accented, noisy, reverberant, far-field, clipped and phone-codec speech — for transcription accuracy and streaming latency.

Technical specifications

Made by
ElevenLabs
Hosted by
ElevenLabs
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Multilingual, VAD, Keyterm biasing

How Scribe v2 Realtime ranks

Full STT dashboard
Time to Final Segment — every measured STT modelMilliseconds · lower is better · 30-day averageEvery STT model ranked on Time to Final Segment, with Scribe v2 Realtime highlighted.
  1. #1STT RT v564 ms
  2. #2Parakeet TDT 0.6B v370 ms
  3. #3STT 183 ms
  4. #4Nova 399 ms
  5. #5Nova 2101 ms
  6. #6Ink 2108 ms
  7. #7Scribe v2 Realtime120 ms
  8. #8Universal 3.5 Pro146 ms
  9. #9Defaultvia Azure155 ms
  10. #10Whisper Large v3180 ms
  11. #11Velma 2 STT Streaming191 ms
  12. #12Grok STT199 ms
Show all 24 models
  1. #13Pulse205 ms
  2. #14Defaultvia Speechmatics223 ms
  3. #15Defaultvia Gradium249 ms
  4. #16resonant-1288 ms
  5. #17Enhanced341 ms
  6. #18Voxtral Mini Transcribe Realtime 2602356 ms
  7. #19GPT Realtime Whisper559 ms
  8. #20GPT-4o mini Transcribe623 ms
  9. #21Solaria 1680 ms
  10. #22GPT-4o Transcribe742 ms
  11. #23Chirp 2811 ms
  12. #24Chirp 3813 ms
median of all models · 202 ms
STT models over the last 30 days, ranked on Time to Final Segment.
#ModelHostTTFSWERTTFTSamples
1STT RT v5Soniox64 ms1533 ms29,322
2Parakeet TDT 0.6B v3Together AI70 ms1210 ms29,161
3STT 1Inworld AI83 ms1468 ms29,280
4Nova 3Deepgram99 ms1434 ms29,290
5Nova 2Deepgram101 ms1436 ms29,145
6Ink 2Cartesia108 ms1812 ms29,288
7Scribe v2 RealtimeElevenLabs120 ms2157 ms29,313
8Universal 3.5 ProAssemblyAI146 ms1030 ms29,280
9DefaultAzure155 ms1792 ms6,682
10Whisper Large v3Together AI180 ms1241 ms29,171
11Velma 2 STT StreamingModulate191 ms1576 ms17,895
12Grok STTxAI199 ms—29,322
13PulseSmallest205 ms2072 ms29,313
14DefaultSpeechmatics223 ms1473 ms29,323
15DefaultGradium249 ms1979 ms28,235
16resonant-1Reson8288 ms—17,066
17EnhancedSpeechmatics341 ms1536 ms29,322
18Voxtral Mini Transcribe Realtime 2602Mistral356 ms1828 ms29,175
19GPT Realtime WhisperOpenAI559 ms1823 ms29,286
20GPT-4o mini TranscribeOpenAI623 ms—29,318
21Solaria 1Gladia680 ms1703 ms26,356
22GPT-4o TranscribeOpenAI742 ms—29,318
23Chirp 2Google811 ms6000 ms29,237
24Chirp 3Google813 ms5999 ms29,319
—FluxDeepgram—1089 ms29,354
—Flux MultilingualDeepgram—1175 ms29,344
—Nemotron 3.5 ASR StreamingTogether AI—1540 ms29,193
—Universal StreamingAssemblyAI—1510 ms29,297
Under-sampled models are excluded; tied models share a place. Dotted WER values split into substitutions, deletions and insertions on hover or tap.

Latency vs accuracy

TTFS vs WEREach point is one measured model · Scribe v2 Realtime highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with Scribe v2 Realtime highlighted.
STT RT v5 — 64 ms, 5.8%Parakeet TDT 0.6B v3 — 70 ms, 11.0%STT 1 — 83 ms, 4.8%Nova 3 — 99 ms, 6.3%Nova 2 — 101 ms, 7.9%Ink 2 — 108 ms, 4.9%Universal 3.5 Pro — 146 ms, 3.2%Default via Azure — 155 ms, 5.3%Whisper Large v3 — 180 ms, 8.2%Velma 2 STT Streaming — 191 ms, 6.4%Grok STT — 199 ms, 4.8%Pulse — 205 ms, 5.3%Default via Speechmatics — 223 ms, 5.5%Default via Gradium — 249 ms, 10.0%resonant-1 — 288 ms, 3.4%Enhanced — 341 ms, 4.3%Voxtral Mini Transcribe Realtime 2602 — 356 ms, 6.2%GPT Realtime Whisper — 559 ms, 5.2%GPT-4o mini Transcribe — 623 ms, 5.1%Solaria 1 — 680 ms, 7.1%GPT-4o Transcribe — 742 ms, 4.7%Chirp 2 — 811 ms, 5.0%Chirp 3 — 813 ms, 4.1%Scribe v2 Realtime — 120 ms, 5.5%

Where the errors come from

WER compositionScribe v2 Realtime's Word Error Rate split by error type · 30-day averageScribe v2 Realtime's WER split into substitutions, deletions and insertions.
  • Scribe v2 Realtime5.5%
SubstitutionsDeletionsInsertions

Averages and tail latency

Averages hide slow outliers — these are the distributions behind each figure.

Scribe v2 Realtime latency distributionMilliseconds · shared axis across metrics · last 30 daysScribe v2 Realtime's latency percentiles per metric: p25–p75 band, p50 tick, whisker to p99.
  • Time to Final Segmentp50 111 ms · p99 297 ms
  • Time to First Tokenp50 2112 ms · p99 3226 ms
band p25–p75 · tick p50 · whisker to p99 with p90 and p95 stops
Average and percentile values per metric, with the number of samples behind each row.
MetricAveragep25p50p75p90p95p99Samples
Time to Final Segment120 ms83 ms111 ms143 ms176 ms199 ms297 ms29,313
Word Error Rate5.5%0.0%0.0%7.1%14.3%22.7%47.6%29,354
Time to First Token2157 ms2055 ms2112 ms2131 ms2216 ms2911 ms3226 ms29,245

Last 30 days

Daily medians from the same measurement runs · gaps are days without qualifying runs.

Time to Final Segment — daily p50Line p50 · band p25–p75 · UTC daysScribe v2 Realtime's daily median Time to Final Segment over the last 30 days.
Scribe v2 Realtime · Jul 31: 119 msScribe v2 Realtime · Aug 1: 94 msScribe v2 Realtime · Aug 2: 90 msScribe v2 Realtime · Aug 3: 120 msScribe v2 Realtime · Aug 4: 118 msScribe v2 Realtime · Aug 5: 128 msScribe v2 Realtime · Aug 6: 122 msScribe v2 Realtime · Aug 7: 124 msScribe v2 Realtime · Aug 8: 97 msScribe v2 Realtime · Aug 9: 87 msScribe v2 Realtime · Aug 10: 95 msScribe v2 Realtime · Aug 11: 86 msScribe v2 Realtime · Aug 12: 107 msScribe v2 Realtime · Aug 13: 114 msScribe v2 Realtime · Aug 14: 96 msScribe v2 Realtime · Aug 15: 85 msScribe v2 Realtime · Aug 16: 71 msScribe v2 Realtime · Aug 17: 116 msScribe v2 Realtime · Aug 18: 130 msScribe v2 Realtime · Aug 19: 109 msScribe v2 Realtime · Aug 20: 103 msScribe v2 Realtime · Aug 21: 101 msScribe v2 Realtime · Aug 22: 102 msScribe v2 Realtime · Aug 23: 88 msScribe v2 Realtime · Aug 24: 140 msScribe v2 Realtime · Aug 25: 133 msScribe v2 Realtime · Aug 26: 139 msScribe v2 Realtime · Aug 27: 121 msScribe v2 Realtime · Aug 28: 132 ms
Word Error Rate — daily averageDaily average · UTC daysScribe v2 Realtime's daily Word Error Rate over the last 30 days.
Scribe v2 Realtime · Jul 31: 6.0%Scribe v2 Realtime · Aug 1: 6.4%Scribe v2 Realtime · Aug 2: 6.8%Scribe v2 Realtime · Aug 3: 5.6%Scribe v2 Realtime · Aug 4: 6.5%Scribe v2 Realtime · Aug 5: 7.2%Scribe v2 Realtime · Aug 6: 5.0%Scribe v2 Realtime · Aug 7: 8.1%Scribe v2 Realtime · Aug 8: 7.4%Scribe v2 Realtime · Aug 9: 6.6%Scribe v2 Realtime · Aug 10: 7.4%Scribe v2 Realtime · Aug 11: 6.8%Scribe v2 Realtime · Aug 12: 6.5%Scribe v2 Realtime · Aug 13: 5.2%Scribe v2 Realtime · Aug 14: 6.4%Scribe v2 Realtime · Aug 15: 6.8%Scribe v2 Realtime · Aug 16: 6.8%Scribe v2 Realtime · Aug 17: 6.0%Scribe v2 Realtime · Aug 18: 8.6%Scribe v2 Realtime · Aug 19: 6.3%Scribe v2 Realtime · Aug 20: 6.7%Scribe v2 Realtime · Aug 21: 5.3%Scribe v2 Realtime · Aug 22: 7.9%Scribe v2 Realtime · Aug 23: 5.1%Scribe v2 Realtime · Aug 24: 7.7%Scribe v2 Realtime · Aug 25: 5.0%Scribe v2 Realtime · Aug 26: 6.0%Scribe v2 Realtime · Aug 27: 6.3%Scribe v2 Realtime · Aug 28: 7.7%

Time to Final Segment by dataset

Scribe v2 Realtime by test conditionMilliseconds · lower is better · best condition firstScribe v2 Realtime's Time to Final Segment on each benchmark dataset.
  1. WildASR accents119 ms
  2. PipeCat (production)119 ms
  3. WildASR phone codec120 ms
  4. WildASR clean121 ms
  5. WildASR reverb122 ms
  6. WildASR noise gaps122 ms
  7. WildASR far-field123 ms
  8. WildASR clipping127 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
LibriSpeech85 ms1
ProductionPipeCat119 ms14,937
AccentsWildASR119 ms1,438
CleanWildASR121 ms5,775
ClippingWildASR127 ms1,442
Far-fieldWildASR123 ms1,441
Noise gapsWildASR122 ms1,440
Phone codecWildASR120 ms1,441
ReverbWildASR122 ms1,398

Strongest condition: LibriSpeech at 85 ms · weakest: WildASR clipping at 127 ms.

Limits of this comparison

Scribe is measured separately from ElevenLabs synthesis models, focusing on transcript errors and streaming timing.

  • The benchmark does not score every language, speaker-labeling mode or event type offered by Scribe.

Official sources

Results are re-measured daily using fixed datasets and reported over a rolling 30-day window. Same datasets, prompts and metric definitions for every model, measured by Coval’s open-source runner. Full methodology on the overview.

Evaluate your own voice agent

Use Coval to test your production configuration, prompts and calls—not only the public benchmark endpoints.

Book a Demo