TTS-2 key stats
- Latency (measured)
- 288 ms1
- Vapi streaming benchmark (50 trials per model) (checked 2026-06-10) Median of 50 sequential live streaming trials, June 2026; includes network RTT from the benchmark machine.
- inworld.ai/blog/realtime-tts-2 (checked 2026-06-10) One voice identity held across more than 100 languages.
- inworld.ai/pricing (checked 2026-06-10) On-demand rate, $25 per 1M characters (cheaper tier than TTS 1.5 Max).
- inworld.ai/blog/realtime-tts-2 (checked 2026-06-10) Research preview.
Background
Realtime TTS-2 is Inworld's frontier voice model, released in May 2026 as a research preview. It conditions on the audio of prior conversation turns, picking up the user's tone, pacing, and emotional state before deciding not just what to say but how to say it. Developers direct it with plain English steering instructions instead of fixed emotion enums, and it holds a single voice identity across more than 100 languages. Teams on TTS-1.5 upgrade by changing one model identifier.
Sources: inworld.ai
At a glance
Conversation aware synthesis, plain English steering tags, Advanced Voice Design personas, and integration partners that include Vapi. In our 50 trial streaming benchmark it returned first audio in a median of 288 ms.
Sources: inworld.ai
Position in the rankings
Standings as of Jun 15, 2026, 08:25 UTC
Frequently asked questions
- How is TTS-2 tested on the Humanness Index™?
- Listeners hear TTS-2 against another model in a blind head to head round, both voices reading the same customer support prompt from the same cloned source voice, and they pick whichever sounds more human. Its Humanness score derives purely from those votes.
- What makes Realtime TTS-2 different?
- It conditions on the audio of prior conversation turns, picking up the user's tone and pacing before deciding how to speak. Developers steer it with plain English instructions, and it holds one voice identity across more than 100 languages.
Find the most human-sounding voice for your agent.
Compare the models in blind tests, read the methodology, or get in touch.
Build a TTS model? Add yours to the Index.
