Google is preparing to introduce new generative models for creating images and videos — Veo 3 and Imagen 4. Their release is expected at the end of May, likely during the annual I/O developer conference. References have appeared to versions “veo-3.0-generate-preview,” “imagen-4.0-generate-preview-05-20,” and “imagen-4.0-ultra-generate-exp-05-20,” indicating a gradual rollout and several capability tiers for different tasks.
Veo will remain focused on video generation, while Imagen will be used for creating photorealistic and stylized images. The “preview” and “ultra” labels in the model names point to different performance options, which may be aimed at creative, commercial, or research user needs.
Imagen 3.5 and Veo 3 are also expected to become available for early testing through Google Labs. Previous versions of these models have already been used to generate media content in products such as NotebookLM and Gemini, allowing users to seamlessly move from text to images and videos in a single environment.
Official details about the new features are still limited, but the transition to the fourth version of Imagen and the third version of Veo indicates improved quality, better coherence of generated sequences, and expanded capabilities for working with different types of content.
