Data read October 27, 2026
Visible delay TTS APIs in 2026: 18 models with a P95 above 293 ms
18 models fall in the Visible delay tier: a P95 above 293 ms on Coval's 30-day measure from the US East. The first is Soniox TTS RT v1.
Tiers are drawn at the widest gaps between consecutive P95 values, so they do not depend on any one model. A model Coval does not measure has no tier.
How the rows are ordered. Ranked models first, in ranking order (speed tier, then P95, ties by median); the 4 models without a rank follow alphabetically.
| Rank | Model | Median (ms) Typical wait before the first sound | P95 (ms) 1 turn in 20 is slower | Languages tested | Listener Elo Artificial Analysis, with 95% interval | Price per 1M characters List price; a range when sources differ |
|---|---|---|---|---|---|---|
| 10 = |  Soniox TTS RT v1 | 258.3 | 319.0 | 3 | n/a | $13 to $14.10 |
| 10 = |  Soniox TTS Real-Time v2 | 264.4 | 324.7 | 10 | 1,180 ±11 | $13 to $14.81 |
| 12 = |  Murf AI Falcon 2 | 171.3 | 363.5 | 0 | 1,154 ±12 | $10 |
| 12 = |  Rime Mist v3 | 275.2 | 359.9 | 0 | n/a | $30 |
| 14 |  Cartesia Sonic 3.6 | 285.8 | 374.4 | 9 | 1,292 ±14 | $39.20 to $50 |
| 15 |  Cartesia Sonic 3.5 | 279.3 | 390.5 | 15 | 1,190 ±11 | $39.20 to $50 |
| 16 = |  Deepdub PhantomZ 4.0 | 183.4 | 430.0 | 0 | n/a | n/a |
| 16 = |  Rime Coda | 317.5 | 428.1 | 1 | 1,054 ±12 | $50 |
| 18 |  SpaceXAI Grok TTS | 329.8 | 473.9 | 11 | 1,155 ±13 | $15 |
| 19 |  Fish Audio S2.1 Pro | 328.1 | 533.0 | 2 | 1,184 ±13 | $15 |
| 20 |  Smallest.ai Lightning V3.1 Pro (Jul 2026) | 360.5 | 703.3 | 0 | 1,169 ±12 | $19.50 |
| 21 |  Alibaba Qwen3 TTS Flash Realtime | 726.8 | 785.8 | 0 | n/a | $13 |
| 22 |  Google Chirp 3: HD | 534.1 | 917.2 | 5 | 1,061 ±12 | $30 |
| 23 |  Fish Audio S2.1 Pro Free | 793.8 | 1,137.9 | 0 | n/a | n/a |
| not ranked |  Deepgram Aura 2 | 263.0 | 525.4 | 3 | n/a | $30 |
| not ranked |  Deepgram Aura 2 En | 527.2 | 985.7 | 0 | n/a | n/a |
| not ranked | Guava Daytona TTS | 450.0 | 584.8 | 0 | n/a | n/a |
| not ranked | Out of Set Airy TTS | 212.3 | 453.9 | 0 | n/a | $3.67 |
Frequently asked questions
- What does the Visible delay speed tier mean?
- A TTS model is in the Visible delay tier when it has a P95 above 293 ms. P95 is the slow turn: one call in twenty is slower.
- Which TTS APIs are in the Visible delay tier?
- Starting with the best-ranked, Soniox TTS RT v1, Soniox TTS Real-Time v2, Murf AI Falcon 2, Rime Mist v3, and Cartesia Sonic 3.6. There are 18 in total.