Skip to content
Inworld AI· inworld·· 16 d agoPickAI score44

Inworld says Realtime TTS-2 Flash starts speaking in 46ms with 93.2% accuracy on hard text on Cekura voice benchmark

Realtime TTS-2 Flash takes 46ms to start speaking. 93.2% accuracy on hard text on @cekuraai’s new independent voice benchmarks.

AI summary

Inworld published Realtime TTS-2 Flash's results on Cekura's new independent voice benchmark: 46ms to start speaking and 93.2% accuracy on hard text. Cekura says the benchmark covers 50+ models from vendors including OpenAI, Google, Nvidia, xAI and Inworld, and allows comparison of naturalness, performance and cost across speech-to-speech, speech-to-text and text-to-speech models and voice agent platforms. Inworld also congratulated the Cekura team on launching the benchmark, saying better evaluation helps everyone build better voice agents.

Why it matters

The 46ms time to first audio and 93.2% accuracy on hard text come from a third-party voice benchmark, giving a quantitative reference point when choosing a voice agent.

Source: Inworld AI · x.com