Use case · Voice agents · read on October 27, 2026
Best text-to-speech for voice agents: 13 TTS models ranked by speed
Agents that hold a live conversation over the phone or the web, where every extra 100 ms between turns is heard. Only the fastest models qualify: this page lists the models whose slow turns (P95) stay under 293 ms.
Recommendation for voice agents
Best for: real-time phone and web voice agents. On this page the shortest wait for the first sound belongs to Fluxions vui (75 ms at P95), then Gradium TTS. One question narrows the choice: does a person wait on the line for the answer? If so, stay in the Headroom and Fast enough tiers.
All 13 models, fastest first
Order applied: speed tier, then P95, ties within 10 ms ordered by median time to first audio; models Coval does not time follow, ordered by listener elo.
| Rank | Model | Speed tier | P95 | Speed (shorter is faster) | Median | Listener Elo | Price per 1M chars | Best for |
|---|---|---|---|---|---|---|---|---|
| 1 | | Headroom | 75 ms | 39.7 ms | Not rated | $9 | Voice agents | |
| 2 |  Gradium TTSGradium | Headroom | 104.3 ms | 53.7 ms | 1,174 | $36 to $57.8 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 3 |  Inworld Realtime TTS-2 FlashInworld | Headroom | 100.9 ms | 58.1 ms | 1,218 | $10.4 to $15 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 4 |  Alibaba Qwen3 TTS FastAlibaba (Qwen) | Headroom | 98.8 ms | 62.9 ms | Not rated | $10 | Voice agents | |
| 5 |  Alibaba Qwen3 TTS 1.7bAlibaba (Qwen) | Fast enough | 136.6 ms | 103.6 ms | Not rated | Not collected | Voice agents | |
| 6 |  SpeechifyAI Simba 3.0Speechify | Fast enough | 149.4 ms | 100.5 ms | 1,116 | $6.6 to $10 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 7 |  SpeechifyAI Simba 3.2Speechify | Fast enough | 161.1 ms | 104.3 ms | 1,249 | $6.6 to $10 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 8 |  Inworld Realtime TTS-2Inworld | Fast enough | 194.7 ms | 149.1 ms | 1,267 | $15 to $25 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 9 |  Palabra TTS v1Palabra | Fast enough | 214.3 ms | 105.1 ms | Not rated | $12 | Voice agents | |
| 10 |  Fish Audio Drama 3 PreviewFish Audio | Fast enough | 209.3 ms | 146 ms | 1,245 | $15 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 11 |  ElevenLabs Eleven v4 TurboElevenLabs | Fast enough | 238.6 ms | 188.9 ms | 1,333 | $40 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 12 |  ElevenLabs Flash v2.5ElevenLabs | Fast enough | 255.1 ms | 191 ms | 1,076 | $40 to $50 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing | |
| 13 |  Deepgram Flux TTSDeepgram | Fast enough | 267 ms | 197.7 ms | 1,052 | $45 | Customer service, Voice agents, Assistants, Entertainment, Knowledge sharing |
The same models, as listeners rate them
Artificial Analysis lets people compare voices blind. The ten highest listener elo scores among these models, highest first, with the 95% interval of each score. Overlapping intervals mean the order is not settled.
| # | Model | Listener Elo | 95% interval | Speed tier | Price per 1M chars |
|---|---|---|---|---|---|
| 1 | ElevenLabs Eleven v4 TurboElevenLabs | 1,333 | ±16 | Fast enough | $40 |
| 2 | Inworld Realtime TTS-2Inworld | 1,267 | ±14 | Fast enough | $15 to $25 |
| 3 | SpeechifyAI Simba 3.2Speechify | 1,249 | ±12 | Fast enough | $6.6 to $10 |
| 4 | Fish Audio Drama 3 PreviewFish Audio | 1,245 | ±14 | Fast enough | $15 |
| 5 | Inworld Realtime TTS-2 FlashInworld | 1,218 | ±13 | Headroom | $10.4 to $15 |
| 6 | Gradium TTSGradium | 1,174 | ±13 | Headroom | $36 to $57.8 |
| 7 | SpeechifyAI Simba 3.0Speechify | 1,116 | ±11 | Fast enough | $6.6 to $10 |
| 8 | ElevenLabs Flash v2.5ElevenLabs | 1,076 | ±11 | Fast enough | $40 to $50 |
| 9 | Deepgram Flux TTSDeepgram | 1,052 | ±13 | Fast enough | $45 |
Frequently asked questions
- What is the best text-to-speech model for voice agents?
- Ranked by speed, Fluxions vui comes first of 13 models with a P95 time to first audio of 75 ms, followed by Gradium TTS. Speed is the first filter; check the listener Elo and price columns before you choose.
- How are the 13 models ordered?
- By speed tier, then by P95 time to first audio from Coval's 30-day window. Two P95 values within 10 ms are a tie, ordered by median time to first audio. Models Coval does not time follow, ordered by listener elo from Artificial Analysis.
- Which models are fast enough for voice agents?
- Fluxions vui, Gradium TTS, Inworld Realtime TTS-2 Flash and Alibaba Qwen3 TTS Fast are in the Headroom tier (P95 under 120 ms) on this page. 9 more are in the Fast enough tier, up to 293 ms.
- Does the price change the order?
- No. Price per million characters is shown in the table as each benchmark prints it, but it never moves a model up or down: speed, voice quality, accuracy and price are read side by side.