Voice AI BenchmarksBy

MODULATESPEECH-TO-TEXT89,565 SAMPLES / 30 DAYSLAST RUN AUG 20, 2026, 23:00 UTC

official resource

Velma 2 STT Streaming speech-to-text benchmarks

Velma 2 STT Streaming is Modulate's real-time transcription model.

Time to Final Segment#11 / 24
191ms
Word Error Rate#19 / 28
6.4%
Time to First Token#14 / 24
1576ms

Overview

Velma is built for live voice applications, with multilingual transcription and speaker diarization.

Velma 2 STT Streaming 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
Modulate
Hosted by
Modulate
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Multilingual, Diarization

How Velma 2 STT Streaming 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 Velma 2 STT Streaming 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 · Velma 2 STT Streaming highlighted · 30-day averagesTime to Final Segment against Word Error Rate for every measured STT model, with Velma 2 STT Streaming 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%Scribe v2 Realtime — 120 ms, 5.5%Universal 3.5 Pro — 146 ms, 3.2%Default via Azure — 155 ms, 5.3%Whisper Large v3 — 180 ms, 8.2%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%Velma 2 STT Streaming — 191 ms, 6.4%

Where the errors come from

WER compositionVelma 2 STT Streaming's Word Error Rate split by error type · 30-day averageVelma 2 STT Streaming's WER split into substitutions, deletions and insertions.
  • Velma 2 STT Streaming6.4%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

Velma 2 STT Streaming latency distributionMilliseconds · shared axis across metrics · last 30 daysVelma 2 STT Streaming's latency percentiles per metric: p25–p75 band, p50 tick, whisker to p99.
  • Time to Final Segmentp50 86 ms · p99 1644 ms
  • Time to First Tokenp50 1573 ms · p99 3380 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 Segment191 ms72 ms86 ms121 ms312 ms576 ms1644 ms17,895
Word Error Rate6.4%0.0%0.0%8.7%16.7%25.0%66.7%17,919
Time to First Token1576 ms1188 ms1573 ms1767 ms2082 ms2362 ms3380 ms17,919

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 daysVelma 2 STT Streaming's daily median Time to Final Segment over the last 30 days.
Velma 2 STT Streaming · Jul 30: 91 msVelma 2 STT Streaming · Jul 31: 85 msVelma 2 STT Streaming · Aug 1: 78 msVelma 2 STT Streaming · Aug 2: 71 msVelma 2 STT Streaming · Aug 3: 91 msVelma 2 STT Streaming · Aug 4: 97 msVelma 2 STT Streaming · Aug 5: 90 msVelma 2 STT Streaming · Aug 6: 109 msVelma 2 STT Streaming · Aug 7: 95 msVelma 2 STT Streaming · Aug 8: 76 msVelma 2 STT Streaming · Aug 9: 83 msVelma 2 STT Streaming · Aug 10: 119 msVelma 2 STT Streaming · Aug 11: 89 msVelma 2 STT Streaming · Aug 12: 92 msVelma 2 STT Streaming · Aug 13: 97 msVelma 2 STT Streaming · Aug 14: 91 msVelma 2 STT Streaming · Aug 15: 79 msVelma 2 STT Streaming · Aug 16: 79 msVelma 2 STT Streaming · Aug 17: 84 msVelma 2 STT Streaming · Aug 18: 117 msVelma 2 STT Streaming · Aug 19: 86 msVelma 2 STT Streaming · Aug 20: 100 ms
Word Error Rate — daily averageDaily average · UTC daysVelma 2 STT Streaming's daily Word Error Rate over the last 30 days.
Velma 2 STT Streaming · Jul 30: 6.6%Velma 2 STT Streaming · Jul 31: 7.2%Velma 2 STT Streaming · Aug 1: 7.1%Velma 2 STT Streaming · Aug 2: 9.0%Velma 2 STT Streaming · Aug 3: 6.5%Velma 2 STT Streaming · Aug 4: 8.5%Velma 2 STT Streaming · Aug 5: 9.3%Velma 2 STT Streaming · Aug 6: 8.5%Velma 2 STT Streaming · Aug 7: 7.9%Velma 2 STT Streaming · Aug 8: 7.9%Velma 2 STT Streaming · Aug 9: 6.5%Velma 2 STT Streaming · Aug 10: 13.6%Velma 2 STT Streaming · Aug 11: 10.7%Velma 2 STT Streaming · Aug 12: 7.1%Velma 2 STT Streaming · Aug 13: 6.3%Velma 2 STT Streaming · Aug 14: 7.8%Velma 2 STT Streaming · Aug 15: 6.6%Velma 2 STT Streaming · Aug 16: 6.4%Velma 2 STT Streaming · Aug 17: 6.8%Velma 2 STT Streaming · Aug 18: 6.8%Velma 2 STT Streaming · Aug 19: 8.3%Velma 2 STT Streaming · Aug 20: 6.4%

Time to Final Segment by dataset

Velma 2 STT Streaming by test conditionMilliseconds · lower is better · best condition firstVelma 2 STT Streaming's Time to Final Segment on each benchmark dataset.
  1. WildASR clean170 ms
  2. WildASR accents180 ms
  3. WildASR far-field182 ms
  4. WildASR clipping184 ms
  5. WildASR phone codec191 ms
  6. WildASR noise gaps194 ms
  7. PipeCat (production)200 ms
  8. WildASR reverb205 ms
Time to Final Segment per dataset, with the number of samples behind each figure.
DatasetTTFSSamples
LibriSpeech699 ms1
ProductionPipeCat200 ms9,318
AccentsWildASR180 ms874
CleanWildASR170 ms3,506
ClippingWildASR184 ms844
Far-fieldWildASR182 ms848
Noise gapsWildASR194 ms842
Phone codecWildASR191 ms853
ReverbWildASR205 ms809

Strongest condition: WildASR clean at 170 ms · weakest: LibriSpeech at 699 ms.

Limits of this comparison

Coval evaluates recognition rather than blending it with Modulate's wider voice-safety and moderation products.

  • The benchmark does not measure moderation, toxicity detection or other analysis products in Modulate's platform.

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