Voice AI BenchmarksBy

FISH AUDIOTEXT-TO-SPEECH7,730 SAMPLES / 30 DAYSLAST RUN OCT 14, 2026, 20:00 UTC

official resource

S1 text-to-speech benchmarks

S1, hosted by Fish Audio (FishAudio, fish.audio), measures mean 390 ms time to first audio (25th of 32) and 1.8% word error rate (17th of 32) among TTS systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

S1 is Fish Audio's earlier text-to-speech generation, still served alongside the newer S2.1 line.

Time to First Audio#25 / 32
390ms
TTFA Network Roundtrip#25 / 32
272ms
TTFA Leading Silence#23 / 32
119ms
Word Error Rate#17 / 32
1.8%

Overview

Fish Audio provides S1 through its cloud API with multilingual, cloning and emotion-control capabilities.

S1 is tested every day on a fixed set of text prompts, measuring how quickly audible speech starts and how intelligible the result is.

Technical specifications

Made by
Fish Audio
Hosted by
Fish Audio
Source
Official API
Licensing
Proprietary
Deployment
Cloud
Region
US
Features
Emotion control, Multilingual, Voice cloning

How S1 ranks

Full TTS dashboard
Time to First Audio — every measured TTS modelMilliseconds · lower is better · 30-day averageEvery TTS model ranked on Time to First Audio, with S1 highlighted.
  1. #1vui59 ms
  2. #2Default63 ms
  3. #3Qwen3 TTS Fast67 ms
  4. #4TTS Flash 291 ms
  5. #5Qwen3 TTS 1.7b107 ms
  6. #6Simba 3.0111 ms
  7. #7Palabra TTS120 ms
  8. #8Simba 3.2126 ms
  9. #9drama3-preview166 ms
  10. #10TTS 2190 ms
  11. #11Flash v2.5192 ms
  12. #12Eleven v4 Turbo201 ms
  13. #13Flux TTS205 ms
  14. #14TTS RT v1254 ms
  15. #15TTS Rt v2256 ms
  16. #16Mist v3287 ms
  17. #17Phantom Z 3.4 conversational289 ms
  18. #18Sonic 3.5293 ms
  19. #19Aura 2297 ms
  20. #20Falcon 2298 ms
  21. #21Sonic 3.6299 ms
  22. #22Coda332 ms
  23. #23S2.1 Pro341 ms
  24. #24Grok TTS362 ms
  25. #25S1390 ms
Show all 32 models
  1. #26Lightning v3.1 Pro392 ms
  2. #27Daytona TTS486 ms
  3. #28Aura 2 En569 ms
  4. #29Chirp 3 HD598 ms
  5. #30Qwen3 TTS Flash Realtime718 ms
  6. #31S2.1 Pro Free849 ms
  7. #32GPT-4o mini TTS2586 ms
median of all models · 288 ms
TTS models over the last 30 days, ranked on Time to First Audio.
#ModelHostTTFAWERSamples
1vuiFluxions59 ms1,762
2Gradium TTS Beta 202609Gradium63 ms1,934
3Qwen3 TTS FastNari Labs67 ms1,934
4TTS Flash 2Inworld AI91 ms1,738
5Qwen3 TTS 1.7bBaseten107 ms1,248
6Simba 3.0SpeechifyAI111 ms1,933
7Palabra TTS v1Palabra120 ms1,594
8Simba 3.2SpeechifyAI126 ms1,930
9Drama 3 Preview Us EastFish Audio166 ms180
10TTS 2Inworld AI190 ms1,738
11Flash v2.5ElevenLabs192 ms1,924
12Eleven v4 TurboElevenLabs201 ms1,244
13Flux TTSDeepgram205 ms1,797
14TTS RT v1Soniox254 ms1,636
15TTS Rt v2Soniox256 ms1,633
16Mist v3Rime287 ms602
17Phantom Z 3.4 conversationalDeepdub289 ms1,669
18Sonic 3.5Cartesia293 ms1,934
19Aura 2Deepgram297 ms1,866
20Falcon 2Murf298 ms1,933
21Sonic 3.6Cartesia299 ms1,934
22CodaRime332 ms602
23S2.1 ProFish Audio341 ms1,934
24Grok TTSSpaceXAI362 ms1,929
25S1Fish Audio390 ms1,933
26Lightning v3.1 ProSmallest392 ms1,934
27Daytona TTSGuava486 ms1,152
28Aura 2 EnCloudflare569 ms1,933
29Chirp 3 HDGoogle598 ms1,934
30Qwen3 TTS Flash RealtimeAlibaba718 ms1,933
31S2.1 Pro FreeFish Audio849 ms1,927
32GPT-4o mini TTSOpenAI2586 ms1,852
Under-sampled models are excluded; tied models share a place. Dotted WER values split into substitutions, deletions and insertions on hover or tap.

Highest relative placement: 17th of 32 on Word Error Rate.

Latency vs accuracy

TTFA vs WEREach point is one measured model · S1 highlighted · 30-day averagesTime to First Audio against Word Error Rate for every measured TTS model, with S1 highlighted.
vui — 59 ms, 2.1%Default — 63 ms, 1.9%Qwen3 TTS Fast — 67 ms, 1.8%TTS Flash 2 — 91 ms, 1.7%Qwen3 TTS 1.7b — 107 ms, 1.8%Simba 3.0 — 111 ms, 1.5%Palabra TTS — 120 ms, 2.6%Simba 3.2 — 126 ms, 1.9%drama3-preview — 166 ms, 2.0%TTS 2 — 190 ms, 1.6%Flash v2.5 — 192 ms, 1.8%Eleven v4 Turbo — 201 ms, 1.7%Flux TTS — 205 ms, 1.7%TTS RT v1 — 254 ms, 2.4%TTS Rt v2 — 256 ms, 2.4%Mist v3 — 287 ms, 2.1%Phantom Z 3.4 conversational — 289 ms, 1.8%Sonic 3.5 — 293 ms, 1.5%Aura 2 — 297 ms, 1.8%Falcon 2 — 298 ms, 1.7%Sonic 3.6 — 299 ms, 1.5%Coda — 332 ms, 9.5%S2.1 Pro — 341 ms, 1.5%Grok TTS — 362 ms, 1.6%Lightning v3.1 Pro — 392 ms, 1.5%Daytona TTS — 486 ms, 1.6%Aura 2 En — 569 ms, 2.1%Chirp 3 HD — 598 ms, 1.5%Qwen3 TTS Flash Realtime — 718 ms, 3.0%S2.1 Pro Free — 849 ms, 1.5%GPT-4o mini TTS — 2586 ms, 1.8%S1 — 390 ms, 1.8%

Where the errors come from

WER compositionS1's Word Error Rate split by error type · 30-day averageS1's WER split into substitutions, deletions and insertions.
  • S11.8%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

S1 latency distributionMilliseconds · shared axis across metrics · last 30 daysS1's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to First Audiop50 380 ms · Max 1929 ms
  • TTFA Network Roundtripp50 259 ms · Max 1923 ms
  • TTFA Leading Silencep50 120 ms · Max 339 ms
band p25–p75 · tick p50 · whisker to max with p90 and p95 stops
Average and percentile values per metric, with the number of samples behind each row.
MetricAveragep25p50p75p90p95MaxSamples
Time to First Audio390 ms316 ms380 ms447 ms487 ms517 ms1929 ms1,933
TTFA Network Roundtrip272 ms243 ms259 ms279 ms297 ms317 ms1923 ms1,933
TTFA Leading Silence119 ms37 ms120 ms190 ms224 ms242 ms339 ms1,933
Word Error Rate1.8%0.0%0.0%2.5%7.9%12.9%62.5%1,931

Last 30 days

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

Time to First Audio — daily p50Line p50 · band p25–p75 · UTC daysS1's daily median Time to First Audio over the last 30 days.
S1 · Sep 23: 365 msS1 · Sep 24: 411 msS1 · Sep 25: 339 msS1 · Sep 26: 388 msS1 · Sep 27: 391 msS1 · Sep 28: 418 msS1 · Sep 29: 404 msS1 · Sep 30: 420 msS1 · Oct 1: 344 msS1 · Oct 2: 384 msS1 · Oct 3: 394 msS1 · Oct 4: 390 msS1 · Oct 5: 390 msS1 · Oct 6: 392 msS1 · Oct 7: 373 msS1 · Oct 8: 427 msS1 · Oct 9: 373 msS1 · Oct 10: 397 msS1 · Oct 11: 384 msS1 · Oct 12: 383 msS1 · Oct 13: 400 msS1 · Oct 14: 391 ms
Word Error Rate — daily averageDaily average · UTC daysS1's daily Word Error Rate over the last 30 days.
S1 · Sep 23: 4.7%S1 · Sep 24: 3.8%S1 · Sep 25: 3.6%S1 · Sep 26: 3.2%S1 · Sep 27: 5.2%S1 · Sep 28: 11.1%S1 · Sep 29: 5.8%S1 · Sep 30: 1.9%S1 · Oct 1: 4.7%S1 · Oct 2: 5.5%S1 · Oct 3: 4.7%S1 · Oct 4: 6.5%S1 · Oct 5: 2.8%S1 · Oct 6: 4.6%S1 · Oct 7: 5.1%S1 · Oct 8: 5.8%S1 · Oct 9: 2.5%S1 · Oct 10: 1.7%S1 · Oct 11: 5.1%S1 · Oct 12: 5.6%S1 · Oct 13: 4.7%S1 · Oct 14: 7.2%

How fast is S1?

On Fish Audio, S1 measures mean 390 ms time to first audio (25th of 32). Last measured 2026-10-14.

How accurate is S1?

On Fish Audio, S1 measures 1.8% word error rate (17th of 32). Last measured 2026-10-14.

Who hosts S1?

S1 is created by Fish Audio and served by Fish Audio. Coval measures each hosted endpoint separately.

Limits of this comparison

S1 receives the same prompts as S2.1 Pro so the two generations can be compared directly.

  • The fixed benchmark voice does not establish clone similarity, emotion quality or performance for every language.

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