Google promotes Gemini 3.1 Flash Image and 3 Pro Image to general availability, adds video-to-image generation

Google moved two native visual generation models to general availability on May 28, 2026: Gemini 3.1 Flash Image (internally branded "Nano Banana 2") and Gemini 3 Pro Image ("Nano Banana Pro"). The GA launch came alongside a new video-to-image generation capability that lets developers generate thumbnails, movie posters, and summary infographics from video files — including public YouTube URLs.

What's new

• GA models:
  • gemini-3.1-flash-image (Nano Banana 2) — speed-optimized visual generation
  • gemini-3-pro-image (Nano Banana Pro) — high-quality, detail-rich image generation
• New capability: Video-to-image generation, available on gemini-3.1-flash-image
  • Input: video file (direct upload or public YouTube URL) + text prompt
  • Output: high-quality thumbnails, cinematic movie posters, or summary infographics
• Preview deprecations: gemini-3.1-flash-image-preview and gemini-3-pro-image-preview shut down June 25, 2026

Context

Google's Gemini 3.x visual model family extends the Gemini API beyond text and code to cover image and video generation through a unified interface. The naming convention mirrors Google's broader model strategy: Flash variants target speed and cost, Pro variants target quality and detail.

Both models had been available in preview before the May 28 GA promotion. The preview variants (-preview suffix) remain accessible briefly but are scheduled for shutdown on June 25 — developers have less than a month to update any code that references the preview model IDs.

The video-to-image feature is the headline addition. Feeding a video as multimodal context to generate a derived still image — for example, automatically producing a thumbnail from a YouTube upload — compresses a multi-step workflow into a single API call. Google's changelog explicitly cites thumbnails, cinematic movie posters, and summary infographics as output types.

Why it matters

GA status matters operationally: it typically brings SLA commitments and stable pricing that preview models don't carry, making production deployments viable without the risk of sudden endpoint changes.

The video-to-image capability is a differentiated addition to the Gemini API. Neither OpenAI's image generation endpoints nor Anthropic's image tools currently offer a native path from a video URL directly to a generated still. For content publishers and media platforms with large video libraries, a thumbnail generation pipeline powered by gemini-3.1-flash-image could substantially reduce manual production work.

The June 25 preview shutdown means teams need to act quickly. Any application that pinned to gemini-3.1-flash-image-preview or gemini-3-pro-image-preview must update model references within four weeks or face errors.