Voice AI BenchmarksBy

FISH AUDIOTEXT-TO-SPEECH456 SAMPLES / 30 DAYSLAST RUN OCT 14, 2026, 03:30 UTC

drama3-preview text-to-speech benchmarks

drama3-preview, hosted by Fish Audio (FishAudio, fish.audio), measures mean 161 ms time to first audio (9th of 32) and 2.0% word error rate (24th of 32) among TTS systems. Results cover the last 30 days. Coval benchmarks hosted endpoints daily using a consistent evaluation methodology. Last measured .

Coval has active text-to-speech measurements for drama3-preview, created by Fish Audio and served through Fish Audio's own API.

Time to First Audio#9 / 32
161ms
TTFA Network Roundtrip#8 / 32
103ms
TTFA Leading Silence#17 / 32
58ms
Word Error Rate#24 / 32
2%

Overview

drama3-preview 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

How drama3-preview 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 drama3-preview highlighted.
  1. #1vui60 ms
  2. #2Default63 ms
  3. #3Qwen3 TTS Fast68 ms
  4. #4TTS Flash 292 ms
  5. #5Qwen3 TTS 1.7b106 ms
  6. #6Simba 3.0111 ms
  7. #7Palabra TTS v1119 ms
  8. #8Simba 3.2126 ms
  9. #9drama3-preview161 ms
  10. #10Flash v2.5191 ms
  11. #11TTS 2192 ms
  12. #12Eleven v4 Turbo201 ms
Show all 32 models
  1. #13Flux TTS205 ms
  2. #14TTS RT v1254 ms
  3. #15TTS Rt v2256 ms
  4. #16Mist v3287 ms
  5. #17Sonic 3.5293 ms
  6. #18Phantom Z 3.4 conversational297 ms
  7. #19Aura 2297 ms
  8. #20Sonic 3.6299 ms
  9. #21Falcon 2304 ms
  10. #22Coda332 ms
  11. #23S2.1 Pro341 ms
  12. #24Grok TTS362 ms
  13. #25S1389 ms
  14. #26Lightning v3.1 Pro392 ms
  15. #27Daytona TTS486 ms
  16. #28Aura 2 En573 ms
  17. #29Chirp 3 HD598 ms
  18. #30Qwen3 TTS Flash Realtime718 ms
  19. #31S2.1 Pro Free850 ms
  20. #32GPT-4o mini TTS2584 ms
median of all models · 290 ms
TTS models over the last 30 days, ranked on Time to First Audio.
#ModelHostTTFAWERSamples
1vuiFluxions60 ms1,696
2Gradium TTS Beta 202609Gradium63 ms1,868
3Qwen3 TTS FastNari Labs68 ms1,868
4TTS Flash 2Inworld AI92 ms1,672
5Qwen3 TTS 1.7bBaseten106 ms1,152
6Simba 3.0SpeechifyAI111 ms1,867
7Palabra TTS v1Palabra119 ms1,535
8Simba 3.2SpeechifyAI126 ms1,864
9Drama 3 Preview Us EastFish Audio161 ms114
10Flash v2.5ElevenLabs191 ms1,859
11TTS 2Inworld AI192 ms1,672
12Eleven v4 TurboElevenLabs201 ms1,178
13Flux TTSDeepgram205 ms1,731
14TTS RT v1Soniox254 ms1,574
15TTS Rt v2Soniox256 ms1,572
16Mist v3Rime287 ms602
17Sonic 3.5Cartesia293 ms1,868
18Phantom Z 3.4 conversationalDeepdub297 ms1,603
19Aura 2Deepgram297 ms1,800
20Sonic 3.6Cartesia299 ms1,868
21Falcon 2Murf304 ms1,867
22CodaRime332 ms602
23S2.1 ProFish Audio341 ms1,868
24Grok TTSSpaceXAI362 ms1,863
25S1Fish Audio389 ms1,867
26Lightning v3.1 ProSmallest392 ms1,868
27Daytona TTSGuava486 ms1,056
28Aura 2 EnCloudflare573 ms1,867
29Chirp 3 HDGoogle598 ms1,868
30Qwen3 TTS Flash RealtimeAlibaba718 ms1,867
31S2.1 Pro FreeFish Audio850 ms1,861
32GPT-4o mini TTSOpenAI2584 ms1,804
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

TTFA vs WEREach point is one measured model · drama3-preview highlighted · 30-day averagesTime to First Audio against Word Error Rate for every measured TTS model, with drama3-preview highlighted.
vui — 60 ms, 2.1%Default — 63 ms, 1.9%Qwen3 TTS Fast — 68 ms, 1.8%TTS Flash 2 — 92 ms, 1.6%Qwen3 TTS 1.7b — 106 ms, 1.7%Simba 3.0 — 111 ms, 1.5%Palabra TTS v1 — 119 ms, 2.6%Simba 3.2 — 126 ms, 1.9%Flash v2.5 — 191 ms, 1.8%TTS 2 — 192 ms, 1.6%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%Sonic 3.5 — 293 ms, 1.5%Phantom Z 3.4 conversational — 297 ms, 1.8%Aura 2 — 297 ms, 1.8%Sonic 3.6 — 299 ms, 1.5%Falcon 2 — 304 ms, 1.7%Coda — 332 ms, 9.5%S2.1 Pro — 341 ms, 1.5%Grok TTS — 362 ms, 1.6%S1 — 389 ms, 1.8%Lightning v3.1 Pro — 392 ms, 1.5%Daytona TTS — 486 ms, 1.6%Aura 2 En — 573 ms, 2.0%Chirp 3 HD — 598 ms, 1.5%Qwen3 TTS Flash Realtime — 718 ms, 3.0%S2.1 Pro Free — 850 ms, 1.5%GPT-4o mini TTS — 2584 ms, 1.8%drama3-preview — 161 ms, 2.0%

Where the errors come from

WER compositiondrama3-preview's Word Error Rate split by error type · 30-day averagedrama3-preview's WER split into substitutions, deletions and insertions.
  • drama3-preview2.0%
SubstitutionsDeletionsInsertions

Averages and tail latency

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

drama3-preview latency distributionMilliseconds · shared axis across metrics · last 30 daysdrama3-preview's latency percentiles per metric: p25–p75 band, p50 tick, whisker to max.
  • Time to First Audiop50 148 ms · Max 584 ms
  • TTFA Network Roundtripp50 96 ms · Max 460 ms
  • TTFA Leading Silencep50 52 ms · Max 472 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 Audio161 ms129 ms148 ms168 ms199 ms236 ms584 ms114
TTFA Network Roundtrip103 ms89 ms96 ms101 ms116 ms145 ms460 ms114
TTFA Leading Silence58 ms37 ms52 ms68 ms79 ms85 ms472 ms114
Word Error Rate2.0%0.0%0.0%2.5%6.8%9.0%35.7%114

How fast is drama3-preview?

On Fish Audio, drama3-preview measures mean 161 ms time to first audio (9th of 32). Last measured 2026-10-14.

How accurate is drama3-preview?

On Fish Audio, drama3-preview measures 2.0% word error rate (24th of 32). Last measured 2026-10-14.

Who hosts drama3-preview?

drama3-preview is created by Fish Audio and served by Fish Audio. Coval measures each hosted endpoint separately.

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