Google has launched Gemini 3.1 Flash Live, a real-time AI audio model designed for natural conversation with reduced latency and more human-like cadence. It outperforms other models in benchmarks for complex tasks, reasoning, and handling interruptions, making AI voices harder to distinguish from humans. The model is rolling out in Google products and will be available for developers to build conversational AI applications.
Background
AI-generated text and audio have historically had identifiable patterns that reveal their machine origin, but rapid improvements in generative AI are making these distinctions increasingly subtle. Real-time conversational AI has struggled with latency and unnatural speech cadence, limiting its usability in fluid human-like interactions.
- Source
- Ars Technica
- Published
- Mar 27, 2026 at 01:44 AM
- Score
- 7.0 / 10