This new AI model hears your tone, senses your mood, and talks back like a real human. Siri could never. - AOL.com

Inworld AI has launched Realtime TTS-2, a text-to-speech voice model designed to infer a speaker's emotional state by analysing tone, pacing, and pitch, and then dynamically adjust its own vocal characteristics in response. The company describes the model as a step toward more natural and responsive human-AI interaction, contrasting it with earlier voice assistants like Siri that lack emotional modulation.

The model processes audio input in real time to detect emotional cues such as frustration, excitement, or calmness, and adapts its voice's pitch, speed, and emphasis accordingly. Inworld AI positions Realtime TTS-2 for applications in customer service, virtual companionship, and gaming, where emotional resonance can enhance user engagement. The company claims the model can respond with appropriate empathy or enthusiasm based on the context.

Industry observers note that emotional voice AI raises questions about user trust and potential manipulation, though Inworld emphasises transparency in its marketing.