The latest build of the Veo3 generative-video model now lets users upload a single still image and prompt the system to create a moving scene in which on-screen characters deliver user-supplied dialogue. Early demonstrations show accurate lip-sync and greater consistency in character appearance across frames, addressing two long-standing obstacles for AI-assisted filmmaking. Creative-software firm Manus said it has integrated the upgraded model into its production workflow, enabling longer shots, richer visuals and natural-sounding speech without additional post-production. Developers and content creators testing the release say the improvements narrow the gap between AI-generated clips and professional studio output, potentially lowering costs for advertising, entertainment and education projects.