benchmarks/text-to-speech by coval
voice AI benchmark · by Coval

Text-to-Speech (TTS) benchmark — by Coval

An independent benchmark of text-to-speech providers on Time to First Audio (TTFA) and Word Error Rate (WER), measured under production-realistic conditions by Coval. Fastest is eleven_flash_v2_5 (212 ms median TTFA), lowest word error rate is octave-2 (3.8%). Mirrored here with attribution; last 7d.

Run by Coval, a voice-AI evaluation platform

This benchmark is produced by Coval, a voice-AI evaluation platform used by teams like Perplexity, Zoom, ServiceNow, and GEICO to test production voice agents. Coval builds testing and evaluation infrastructure for voice AI — it does not sell TTSmodels of its own — so it measures every provider as a neutral third party rather than benchmarking its own product. Every number here is Coval's, measured on a pinned dataset under production-realistic conditions; the methodology, runner code, and inputs are open-source (Apache-2.0) and reproducible by re-running the suite. Openbenchmarks mirrors these results daily with attribution.

synced from Coval2026-07-12 01:18 UTC · run 4109 · dataset 82cfb82d84c6 · methodology → · benchmarks.coval.ai → · open-source runner — re-run it yourself →

Text-to-Speech models ranked by Time to First Audio

Every model runs the same pinned dataset under the same conditions. TTFA (median) is the headline latency; WER is accuracy (lower is better on both). Ranked fastest-first, last 7d:

#ModelTTFA medianTTFA p95WERSamples
1eleven_flash_v2_5
Elevenlabs
212 ms2,849 ms5.5%3,360
2inworld-tts-1.5-mini
Inworld
228 ms480 ms6.0%3,360
3neural
Azure
235 ms295 ms4.3%990
4s2.1-pro
Fishaudio
239 ms447 ms5.7%990
5inworld-tts-1.5-max
Inworld
250 ms429 ms5.7%3,360
6mistv3
Rime
259 ms291 ms6.3%3,359
7sonic-3.5
Cartesia
276 ms351 ms5.7%3,357
8dragon-hd-latest
Azure
286 ms381 ms4.9%990
9tts-rt-v1
Soniox
293 ms330 ms5.0%3,357
10coda
Rime
301 ms369 ms7.3%3,335
11aura-2-thalia-en
Deepgram
329 ms538 ms7.7%3,354
12default
Gradium
331 ms407 ms3.9%3,360
13arcana
Rime
334 ms380 ms7.6%3,360
14grok-tts
Xai
366 ms464 ms4.0%3,354
15s1
Fishaudio
440 ms621 ms5.5%990
16chirp-3-hd
Google
452 ms815 ms5.8%990
17sonic-3
Cartesia
456 ms808 ms6.2%3,356
18eleven_v3
Elevenlabs
589 ms3,198 ms5.4%1,009
19octave-2
Hume
592 ms805 ms3.8%3,344
20octave-tts
Hume
673 ms945 ms4.2%3,346
21qwen3-tts-flash-realtime
Alibaba
689 ms756 ms9.1%990
22gpt-4o-mini-tts
Openai
812 ms1,604 ms4.8%3,074
23lightning_v3.1_pro
Smallest
939 ms1,132 ms7.5%3,358
24eleven_multilingual_v2
Elevenlabs
1,267 ms1,637 ms4.4%3,081

Numbers are point-in-time against Coval's pinned dataset and refresh continuously — they don't generalize indefinitely. Full charts, distributions, and windows (24h / 7d / 30d) at benchmarks.coval.ai →