Descriptions:
Sam Witteveen reviews Google’s Gemini 3.1 Flash image model — nicknamed “Nano Banana 2” in his series — positioning it as a faster, cheaper alternative to the Gemini 3 Pro image model (“Nano Banana Pro”) that frequently approaches Pro-level quality at a lower price point.
The video walks through hands-on demonstrations of the model’s strongest capabilities: precise text rendering in generated images, multilingual text translation applied to in-scene signage (tested in Italian and Thai), and multi-turn conversational editing. Witteveen runs the same prompts through both Nano Banana 2 and Nano Banana Pro side-by-side, noting where the Flash model falls short on spatial composition while holding its own on text accuracy and instruction following.
Two features stand out as upgrades over the Pro model: improved character consistency with support for up to 14 reference images (useful for product photography and branded content), and a new image search grounding tool not available in the Pro version. The model also retains full aspect ratio support, dual thinking levels, and 512px minimum resolution. Rollout is planned across the Gemini app, Google AI Studio, and Google Cloud Vertex AI, with pricing sitting between the older Gemini 2.5 Flash and the Pro tier.
📺 Source: Sam Witteveen · Published February 26, 2026
🏷️ Format: Review
