Earlier this year, Google introduced their innovative video generation model, Veo, and the latest image generation model, Imagen 3. It's been an exhilarating journey watching creators, from YouTube influencers to enterprise customers, leverage these models to transform their creative visions into reality. Many are exploring Veo and Imagen to innovate video backgrounds for YouTube Shorts, enhance creative workflows on Vertex AI, and narrate stories through VideoFX and ImageFX. With the collaborative efforts of filmmakers and enterprises, we continue to elevate these state-of-the-art video and image generation technologies.
Today marks the introduction of Google's advanced video model, Veo 2, achieving cutting-edge results. Veo 2 is accessible via VideoFX and their Labs experiment, Whisk, offering high-quality videos across diverse subjects and styles. Human raters have judged Veo 2 to deliver state-of-the-art outcomes compared to other leading models. Its enhanced understanding of real-world physics and the subtleties of human movement ensures an improvement in detail and realism.
Veo 2's comprehension of the cinematographic language is unparalleled. Whether you desire a specific genre, lens, or cinematic effect, Veo 2 delivers, supporting up to 4K resolutions and extended video lengths. Request a low-angle shot gliding through a scene or a close-up of a scientist, and Veo 2 achieves it, from wide-angle shots with an “18mm lens” to focused subjects using a "shallow depth of field."
Additionally, Veo 2 minimizes the typical video model “hallucinations” such as extra fingers, ensuring a more realistic output. Guided by our commitment to safety, our measured rollout of Veo 2 to platforms like VideoFX, YouTube, and Vertex AI helps us enhance the model, integrating SynthID watermarks to identify AI-generated content and curb misinformation.
Experience Veo 2's new capabilities today in Google Labs' VideoFX, with plans for expansion to YouTube Shorts and other products next year. Sign up for the waitlist at Google Labs to access this transformative tool.
The Imagen 3 model has also seen significant improvements, offering brighter, better-composed images in diverse art styles, from photorealism to anime. This model's ability to adhere to prompts and render intricate details sets it apart, having achieved state-of-the-art results in human-rated side-by-side comparisons with other top models.
Today, Imagen 3 rolls out globally in ImageFX to over 100 countries. Visit ImageFX to begin using this powerful image generation tool, a testament to our commitment to elevating creative possibilities.
Google's newest Google Labs experiment, Whisk, introduces an exciting way to visualize ideas using images. By inputting or creating images that convey your desired subject, scene, and style, Whisk allows you to remix these elements creatively. Create anything from a digital plushie to an enamel pin or sticker with unique twists.
Whisk leverages the latest Imagen 3 model combined with Gemini’s visual understanding. The Gemini model crafts detailed captions of your images, feeding them into Imagen 3, facilitating novel ways to remix subjects, scenes, and styles.
The advancements in video and image generation with Veo 2 and Imagen 3 mark a significant milestone in digital creativity. These models, supported by tools like Google Labs' Whisk, unlock limitless creative potential for users globally. As Google continue to evolve these technologies, we invite you to explore and innovate using these powerful tools, making creativity more accessible than ever before.
