Inworld launches Realtime TTS-2 with multi-turn context, sub-100ms latency at $12.50/M chars
SucceededMind · x · 2026-09-03
Inworld AI has launched Realtime TTS-2, reframing voice synthesis as a stateful session rather than one-shot "read this sentence" TTS. Key specs: multi-turn audio context with natural-language steering, sub-100ms latency, and $12.50 per million characters — the company claims it outperforms models 20x its cost and now ranks #1 on Artificial Analysis. The author argues stateless TTS has commoditized, but session-aware realtime voice is far harder to commoditize, meaning voice agent architectures just shifted fundamentally.
Related event: Inworld's Realtime TTS-2 Launches and Tops Voice Leaderboards(7 posts)→
More from Multimodal
- One prompt turns Claude Code into a full motion-graphics video pipeline with Remotion — aziz4ai · 2026-10-18
- Flux 3 Text-to-Image Can't Make Realistic Faces at All, Reddit User Reports — EstablishmentFun6629 · 2026-10-18
- MiniMax H3 Ref-to-Video: 15s Shot from 3 Images + Vocals, 1088p Latent Refinement — michael_goldenhour · 2026-10-18
- Dev builds fully local AI tool generating HD 3D textures with Qwen, may sell for $5 — TinfoilTricorn · 2026-10-18
- 14-Agent AI Film Pipeline on One RTX 5090: Fixing Cut Defects Between Clips — developervkmp · 2026-10-18
- 4MP Text-to-Image Comparison: Qwen Image 2.1 Turbo vs MOSS vs Krea 2 vs Flux 3 Klein — FORNAX_460 · 2026-10-18