Thursday, Oct 8 | --:--
Back to home

Gemini Omni Flash and Nano Banana 2 Lite: Media Stack, Not Another Text Frontier GA

Same day Anthropic restored Fable and shipped Sonnet 5, Google opened Omni Flash video preview and Nano Banana 2 Lite image GA—cheap high-volume images plus API-native conversational video for builders, not a pure text frontier launch.

Times of AI Desk 5 min read Mountain View, CA View as Markdown
Cover illustration for Gemini Omni Flash and Nano Banana 2 Lite: Media Stack, Not Another Text Frontier GA

June 30’s Google answer to Anthropic’s Fable restore and Sonnet 5 day was a media developer stack: cheap high-volume images plus API-native conversational video—not another pure text frontier GA.

Google DeepMind dual-shipped Gemini Omni Flash (gemini-omni-flash-preview) for video generation and conversational editing—newly in Google AI Studio, the Gemini API, Enterprise Agent Platform, Gemini app, and Google Flow—and Nano Banana 2 Lite (gemini-3.1-flash-lite-image), the speed/cost tier of the Nano Banana image family, GA for developers and rolling into Search AI Mode, Gemini app, Photos, Ads, and more. Omni Flash video is priced at $0.10/sec (aligned with Veo 3.1 Fast); Nano Banana 2 Lite targets ~$0.034 per 1K-resolution image and ~4s text-to-image latency.

What shipped

From Google’s Keyword primary (“Start building with Nano Banana 2 Lite and Gemini Omni Flash”) and Cloud/API changelogs:

Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image)

  • Fastest, most cost-efficient image model in the Nano Banana family; recommended upgrade from legacy Nano Banana (gemini-2.5-flash-image).
  • Strengths: latency, throughput, prompt adherence, character consistency, legible in-image text.
  • Surfaces: AI Studio, Gemini API, Agent Platform; consumer rollouts (Search AI Mode, Gemini app, NotebookLM, Photos, Stitch, Flow, Ads).

Gemini Omni Flash

  • Multimodal video generation/editing from text, image, and video inputs; conversational refine loops; Gemini world knowledge for scene construction; text/action sync via prompting.
  • Public preview for developers (first broad API access after I/O intro).
  • Limits at launch: ~10-second clips (longer coming); API gaps on audio references, scene extension, and short video refs; character consistency still improving on pans/scene changes.
  • SynthID watermarking on both models; verification paths via Gemini app/Chrome/Search.

Google demos chaining image → animate with Omni (e.g. Anywhere, Space Lift, Omni product studio) via the Interactions API for multi-turn sessions (up to three sequential edits).

Limits

  • Omni Flash clip length ~10s at launch; longer promised, not day-one.
  • Nano Banana 2 Lite pricing (~$0.034/1K) is vendor list economics—volume contracts differ.
  • Gemini 3.5 Pro GA was still delayed into July; this ship does not close that gap.

Sources

  • Google Keyword: “Start building with Nano Banana 2 Lite and Gemini Omni Flash” (June 30, 2026).
  • Google Cloud Blog same-day product post; Gemini API changelog (June 30, 2026).
  • Google DeepMind model pages for Omni / Nano Banana family.

Prior Coverage

Earlier Times of AI reporting on this thread.

Scroll to continue reading