Google Vids brings Gemini Omni and AI avatars to video creation

Gemini Omni turning Google Vids into a prompt‑driven video editor with personal avatars is another step toward AI agents owning the entire creative stack, not just text. The system doesn’t just generate b‑roll; it can synthesize you as a presenter, edit lighting and backgrounds conversationally, and output clips that are automatically watermarked. That combination of generative power plus provenance tooling is exactly where frontier model vendors want to be as regulators start probing deepfake harms.

From a competition standpoint, Google is trying to ensure that as text‑to‑video and avatar tech mature, those workflows happen inside its subscription and Workspace universe rather than in standalone tools like Pika or Runway. For the race to AGI, the interesting piece is not the avatars themselves but the infrastructure: Omni’s ability to maintain character consistency, remember scene context, and follow multi‑step edit instructions is essentially a long‑horizon agent running over video. Scaling those capabilities will feed back into more capable planning and reasoning models well beyond video editing.