Character.AI introduces TalkingMachines, a new AI model that can generate interactive videos

Character.AI’s new feature is powered by a Diffusion Transformer (DiT), which utilises a technique called asymmetric knowledge distillation that enables the conversion of a “high-quality, bidirectional video model into a blazing-fast, real-time generator.” The company says its new model listens to audio and then animates parts of the character like mouth, head and eyes, all in sync with every word, pause and intonation without compromising on things like consistency, image quality, style and expressiveness.

https://images.indianexpress.com/2025/07/intro.mp4?_=2

For audio, Character.AI is using a custom-built 1.2B parameter audio module that captures both speech and silence, with the company claiming that it can achieve “infinite-length generation with no quality degradation over time.” The company goes on to say that its new AI model supports a variety of styles, like photorealistic humans to anime to 3D avatars and builds on the core infrastructure for role-playing, storytelling and interactive world-building.

Character.AI has been constantly adding new features like a new image-to-video generator called AvatarFX, Scenes and Streams. Following OpenAI’s advanced voice mode, the startup even added a call feature that allows users to engage in voice conversations with the character of their choice to increase engagement.

Last year, the Google-owned startup was sued by the mother of a 14-year-old kid in Florida who claimed that a Character.AI chatbot encouraged her son to kill himself. Since then, the company has introduced new supervision tools to ensure the online safety of users under 18.

Curated For You

Kerala Local Body Election Results 2025 LIVE Updates: Congress-led UDF wins big; NDA registers historic Thiruvananthapuram victory

Geminid meteor shower peaks tonight with bright streaks over India: What to know

India vs South Africa 3rd T20 Match: Date, Time, Squad, Predicted Playing XI, Other Details