Uploaded August 2025 | Updated September 2026, 1 hour ago

---

OpenAI just dropped a game-changer for AI voice apps

Here's what's new (and worthwhile) 👇

• A brand-new speech-to-speech model powering a faster, more expressive realtime voice API

• Key upgrades: ultra-low latency + human-like expressiveness → feels like you’re talking to a real person

• Handles emotional cues, accents, and delivery (slow, fast, empathetic, playful, etc.)

• Supports multilingual conversations that adapt live, even mid-sentence

• Async function calling → it keeps talking naturally while tools run in the background (no awkward waiting)

• Better at following instructions (e.g. don't reply to money-related queries)

• Integrates with remote MCP servers, images, and phone calling

• Fine-tuned with real customer feedback → stronger for support, education, and personal assistants

Why it matters?

This isn’t just better text-to-speech. It’s an agent-ready voice model (speech to speech directly) that makes apps sound natural, adaptive, and useful in real time.

I’m Louis-François — PhD dropout, now CTO & co-founder at Towards AI. Follow me for tomorrow’s no-BS AI roundup 🚀

#ai #openai #voiceai #short