Home / Technology / Google's Genie 4: Real-time AI worlds with sound
Google's Genie 4: Real-time AI worlds with sound
27 Oct
Summary
- New AI model simulates interactive environments with generated audio.
- Simulations can now last up to two minutes, doubling previous limits.
- Technology aims to create safe training environments and aid creative tools.
Google DeepMind has launched Genie 4, a groundbreaking AI model that simulates interactive environments in real-time, now generating both visuals and audio. This new model allows users to direct scenes through simple text commands, modifying camera angles, weather, and characters. Ambient sounds, music, and dialogue are generated dynamically as the simulation unfolds.
Building upon Google's Gemini family, Genie 4 offers simulations up to two minutes long, a doubling of the previous Genie 3 capability. The technology is poised to create safe training grounds for robots and agents, while also serving as a powerful tool for filmmaking, journalism, gaming, and manufacturing.
Despite its advancements, limitations persist. Details in longer simulations may become unstable, and achieving perfect consistency over two minutes remains a challenge. Google acknowledges that ensuring accurate audio and legible in-world text are areas for future development. The company plans to roll out Genie 4 to select testers and partners in the coming weeks.
