245.6 ms P50 time-to-first-audio on Coval; ranked 2nd (tied) of 112 in the TTS for Voice Agents 2026 index.
Benchmark summary: 73.7/100, ranked 2nd (tied) of 112 in the TTS for Voice Agents 2026 index (Leader). Strongest criterion: voice quality, 17.5 of 20 points.
Soniox TTS Real-Time v2 is a proprietary text-to-speech (TTS) API from Soniox, measured by three independent benchmarks: Artificial Analysis, Coval and Speko.
Overall score
73.7 /100
Data checked on September 30, 2026
Updated after each benchmark capture
Every point comes from a public benchmark: sources · methodology
Soniox TTS Real-Time v2 ranks 2nd of 112 in the 2026 text-to-speech (TTS) API benchmark, with a score of 73.7/100.
On Coval's 30-day window ending October 1, 2026 (10,104 runs), Soniox TTS Real-Time v2 has a P50 (median) time-to-first-audio (TTFA) of 245.6 ms, 11th lowest of the 31 models Coval measures. Its P99, the slowest 1% of turns, is 343.4 ms (10th lowest of the 31 models Coval measures). Latency consistency: a standard deviation of 49.8 ms (9th lowest of the 31 models Coval measures). Mean time to first byte (TTFB) is 177.7 ms. Speko measures a P50 TTFA of 381 ms from US East (n=30), dated August 13, 2026: 22nd lowest of the 29 models Speko measures.
Coval measures an accuracy (word error rate) of 4.1% over 10,084 samples on the same window, 4th lowest of the 31 models Coval measures. Speko rates its pronunciation robustness on numbers, dates and currency at 0.49 on a 0-to-1 scale (8th highest of the 28 models Speko rates).
In the Artificial Analysis Speech Arena (capture of October 1, 2026), Soniox TTS Real-Time v2 has an Elo of 1,175 (±12, 95% confidence) from 4,776 blind comparisons, 15th highest of the 92 models Artificial Analysis rates. Speko rates its naturalness at 1,606 Elo from blind A/B votes (5th highest of the 41 models Speko rates).
Artificial Analysis lists Soniox TTS Real-Time v2 at $14.2 per 1M characters (18th lowest of the 78 models Artificial Analysis lists a price for). Speko lists a cost of about $13 per 1M characters.
Speko changed its pronunciation robustness between the captures of September 30, 2026 and October 1, 2026 without a new measurement date (0.78, then 0.49): the score uses the average, 0.635. The value shown is the latest.
Tested in Spanish, German, French, Japanese, Chinese, Arabic, Filipino, Norwegian and Thai by Artificial Analysis or Speko. A language missing here was not tested, which does not mean it is unsupported.
Alibaba Qwen3 TTS Fast is an open-weights text-to-speech (TTS) API from Alibaba (Qwen), served by Nari Labs, measured by two independent benchmarks: Coval and Speko.
Gradium TTS Beta is a proprietary text-to-speech (TTS) API from Gradium, measured by two independent benchmarks: Coval and Speko.
Inworld Realtime TTS-2 is a proprietary text-to-speech (TTS) API from Inworld, measured by three independent benchmarks: Artificial Analysis, Coval and Speko.