Humanness Index™ · Provider
Inworld builds voice models for interactive agents, and its TTS line climbed third party speech arenas through 2025, with the 1.5 generation reaching the top of the Artificial Analysis Speech Arena ahead of Google and ElevenLabs.
Standings as of Jun 13, 2026, 01:14 UTC
| Rank | Model | Humanness | Latency | Languages | Price / 1M chars |
|---|---|---|---|---|---|
| #11 | TTS-1.5-max | 65 | 337 ms | 15 | $35 |
| #9 | TTS-2 | 69 | 288 ms | 100+ | $25 |
Inworld builds voice models for interactive agents, and its TTS line climbed third party speech arenas through 2025, with the 1.5 generation reaching the top of the Artificial Analysis Speech Arena ahead of Google and ElevenLabs. Integration partners for its realtime models include Vapi.
Sources: inworld.ai
The 1.5 generation targets the quality and speed balance most production agents need, with a mini sibling for the lowest latency. Realtime TTS-2, released as a research preview in May 2026, conditions on the audio of prior conversation turns and holds a single voice identity across more than 100 languages.
Sources: docs.inworld.ai
The benchmark is open source. Suggest a model, read the methodology, or ask us to put your voice in the arena.
