Video models often “hallucinate” unwanted details, such as extra fingers or unexpected objects, but Veo 2 generates these less frequently, making the output more realistic.
Our commitment to safety and responsible development guides Veo 2. We’ve been purposefully measured in expanding the availability of Veo, allowing us to identify, understand, and improve the quality and safety of our models as we slowly deploy them across VideoFX, YouTube, and Vertex AI.
Like other image and video generation models, Veo 2 output includes an invisible SynthID watermark to identify it as AI-generated, reducing the possibility of misinformation or misattribution.
Today, we’re introducing new Veo 2 capabilities to the Google Labs video generation tool VideoFX, expanding the number of users who can access it. Visit Google Labs to join the waitlist. Next year, we plan to expand Veo 2 to YouTube Shorts and other products.
