Google Releases Gemini Omni Flash API for Enterprise Video Production

Google has made its Gemini Omni Flash AI model available to enterprise customers through an API. This launch enables the creation and editing of video content entirely through conversation, a capability previously introduced to consumers, now targeting the acceleration of internal corporate video production.

Traditionally, producing enterprise videos like 90-second training modules or product explainers has been a complex process involving detailed briefs, production crews, shoots, editing, and revisions. Even minor text changes could necessitate restarting the entire workflow, making video production slow and expensive. Gemini Omni Flash aims to reshape this paradigm by offering a faster, more cost-effective solution.

The new API allows for the production of intricate videos by integrating tools previously used separately. The model can generate video from text, images, and existing video clips, delivering a finished clip with synchronized audio. Users can also edit existing videos by providing conversational instructions, streamlining the post-production process.

The technology supports multimodal inputs, such as reference images and existing video clips, which the model incorporates into the final output. It also understands physical scene behaviors, like lighting and rain effects, and can manage the insertion or alteration of text and logos within videos. However, users are still advised to human-review the generated content before publishing.

Google states current limitations include a maximum clip length of 10 seconds and a 720p resolution. Key features for enterprises also include built-in watermarking (SynthID), content provenance tools (C2PA), and an AI-generated content detection API. Google has also implemented restrictions, such as prohibiting the use of still photos and audio to create deepfakes.