Sparkify combines Gemini and Veo 3 to power Google’s latest AI video experiment

Google continues to develop its Sparkify experiment, first revealed at Google I/O and still available only via a waitlist. Sparkify is positioned as a standalone tool rather than simply an extension of earlier projects like Google Illuminate, where similar AI-generated educational videos first surfaced. The interface provides three main sections: Explore, Generate, and My Library.

The Explore page features themed short videos, typically about two minutes long, such as the origins of drinks or quirky landmark tours, all created with AI and covering various visual styles like cartoon, origami, anime, and realistic. Videos can be vertical or horizontal, closely resembling formats found on TikTok or Instagram Reels. Each video combines sound and visuals to tell a concise story or explain a topic, and the consistent use of certain characters across multiple videos suggests a possible prompt-driven or template-based approach for character persistence.

The core of Sparkify relies on Google’s latest AI models, Gemini and Veo. The inclusion of Gemini hints at a more sophisticated pipeline, likely handling content assembly or narrative structure, while Veo powers the actual video generation. When trying out the generation feature, users are presented with controls to set the topic, visual style, and mood. Options include Felt, Clay, Origami, 3D cartoon, Anime, and Realistic, along with moods such as delightful, academic, or funny. At this stage, though, the actual output is a static demo video, not yet responsive to user prompts, leaving key questions about prompt effectiveness and model constraints unanswered.

Given Sparkify’s output length and quality, there’s speculation that it generates shorter clips and then stitches them together, perhaps where Gemini’s orchestration comes in. This approach would make it possible to create longer, cohesive stories while staying within current generative AI limitations for video duration. The consistency of character design across videos supports the idea that creators could develop branded personalities or themes for social media, powering new types of AI-driven content channels.

For now, Sparkify remains an invite-only experiment with no public release timeline. Its current status reflects Google’s ongoing focus on AI-driven media tools, offering a glimpse at how Gemini and Veo might be leveraged to enable quick, thematic video creation for education or entertainment. TestingCatalog is monitoring how these AI-generated videos perform on platforms like TikTok, as Google continues to fine-tune the experience before a wider rollout.