Skip to content
Hacker News · AI·· 11 h agoAI score60

Inworld launches Realtime TTS-2 and TTS-2 Flash with first audio as low as 50 ms

Inworld TTS looks not bad at all

AI summary

Inworld released two realtime text-to-speech models, Realtime TTS-2 and Realtime TTS-2 Flash, with first audio as low as 90 ms and 50 ms respectively. Realtime TTS-2 covers about 500 languages and dialects and accepts plain-language delivery instructions, while Flash trades expressive steering for latency and cost, priced from $7 per million characters on the Growth plan. Inworld cites Coval benchmarks putting Flash up to 3x sooner to first audio than ElevenLabs and up to 35x sooner than OpenAI GPT-4o mini TTS, and offers an OpenAI-compatible endpoint plus an ElevenLabs migration tool.

Source:Hacker News · AI · inworld.ai