Google adds video and image models to API
Gemini Omni Flash brings conversational video editing at $0.10/sec to the API; Nano Banana 2 Lite adds fast image generation at $0.034 per 1K-resolution image.
Google released two new models on June 30: Gemini Omni Flash and Nano Banana 2 Lite.
Gemini Omni Flash (gemini-omni-flash-preview) is a video generation model focused on conversational editing. You feed it text, images, or existing video and tell it to swap characters, relight a scene, or change an angle in plain language. It currently generates 10-second clips, with longer durations on the roadmap. Pricing is $0.10 per second of video output, matching Veo 3.1 Fast. It is in public preview in the Gemini API, Google AI Studio, and the Enterprise Agent Platform, and is also rolling out in the Gemini app and Google Flow. Known API limitations: audio references and scene extensions above 3 seconds are not yet supported.
Nano Banana 2 Lite (gemini-3.1-flash-lite-image) is a text-to-image model optimized for speed. It generates images at 4 seconds per output at $0.034 per 1,000-resolution image and is rolling to the API, AI Studio, AI Mode in Search, Gemini app, Photos, NotebookLM, and Google Ads.
What this means for builders
Until today, affordable AI video generation via API required piecing together third-party providers. Google now offers a direct path with predictable per-second billing. Omni Flash is not the most capable video model Google has (that is Veo 3), but it is the cheapest way to add video editing into an API workflow.
For image generation, Nano Banana 2 Lite competes on latency rather than quality. At 4 seconds and roughly $0.034 per image it fits use cases where speed matters more than photorealism, such as real-time personalization and ad variations.
Both models support chaining via the Interactions API, keeping session context across multi-turn edits, which matters if you are building anything iterative. Check Google’s earlier Gemini 3.5 Flash Computer Use brief for context on how Google has been extending Gemini’s API capabilities in recent months.