Google releases Gemini Omni Flash: an AI creation model that can generate new videos using video + text

Core conclusion

On May 23, 2026, Google officially released the first version of the Gemini Omni series of models-Omni Flash after the I/O conference. This is an "anything-to-anything" generative AI model currently focused on video generation and has been integrated into Google's AI video editing platform Flow. Compared with the previous generation Veo, Omni Flash supports generating new content based on video clips + text prompts at the same time, and can better maintain character consistency.

Key Points

• Time of Event: May 23, 2026
• Influenced objects: AI video creators, content automation teams, e-commerce and marketers who use AI to generate materials
• Core Change: Upgraded from "Text→Video" to "Video+Text→New Video", the consistency of roles has been greatly improved, and the content creation process may be restructured as a result.

Background and trigger events

Google exhibited the Gemini Omni series of models at the 2026 I/O conference, which is positioned as a unified generative model of "all input → all output" (anything-to-anything). Omni Flash is the first publicly released version, powered by Google's AI video creation and editing platform Flow.

According to actual testing by The Verge reporter Allison Johnson, Omni Flash can upload an existing video clip and then generate new video content with a text prompt. Google claims that the Omni model incorporates more "real-world knowledge" when generating videos and can better maintain the consistency of characters and objects from frame to frame in the video.

The actual test results showed a "confusing mixture" - some scenes were stunning, while others were still unstable.

Google video generation model evolution

Impact on content creation workflow

Positive change

• Material Reuse: Existing video materials can be used as the basis for AI generation, significantly reducing cold start costs.
• Character Consistency: In AI-generated videos, the visual characteristics of characters/objects are more stable and no longer "face-changing" every frame.
• Simplified Editing: Complete the entire process from generation to post-editing directly within the Flow platform

Current limitations

• The output is still video, and multi-modal output such as "text → 3D/audio" has not yet been implemented (the future Omni flagship version will complete it)
• Although character consistency has been improved, "floating" and deformation may still occur in complex scenes
• It requires hands-on testing to determine whether it meets commercially available standards.

Adaptation suggestions

• Content creators can consider incorporating Flow + Omni Flash into the short video production pipeline, starting with a low-cost solution for material reuse
• E-commerce product videos: Use existing product display video clips as seeds to let AI generate variations from different angles
• Watch out for the release of a full-featured version of Omni - "anything-to-anything", once mature, will revolutionize content workflows
• While retaining Veo as an alternative model, Omni Flash is not better than the previous generation in all scenarios

Related extended information

• The Verge ：Google Omni ""
• Google I/O 2026 100

Tool entry (trigger tool floating card)

Tool names that appear naturally in the text include: Gemini, Google, OpenAI, ChatGPT, DeepSeek, Claude. These terms will match the maintained tools library on the platform side, and the hover-card will be displayed on the front end.

Internal link guidance

• Want to use AI to automate content production? Watch: How to build an AI content automated distribution system with n8n + ChatGPT: a complete 30-minute tutorial
• Real case: He earns over 10,000 per month by relying on AI code review + specification-driven development: a practical review of a freelance developer