Google launched Nano Banana 2 Lite, a faster and cheaper text-to-image model. While it excels in speed and cost-effectiveness, especially for text generation and spatial awareness, it shows a significant drop in realism and prompt adherence for detailed photographic outputs compared to Nano Banana 2.
米Googleは、画像生成AI「Nano Banana 2 Lite」と動画生成・編集AI「Gemini Omni Flash」を発表。Nano Banana 2 Liteは高速・低コストで、Gemini Omni Flashはテキスト・画像・動画入力から動画生成や会話形式での編集が可能。両モデルの連携活用も紹介された。
Google DeepMind's new Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) offers a balance of speed and quality for AI image generation, available across Google's ecosystem. It's faster and cheaper than previous models, ideal for rapid prototyping, though it has limitations with text and consistency.
Google launched Nano Banana 2 Lite, an AI video and image generator that is faster and more affordable than previous versions. It can produce images in four seconds for $0.034 per 1,000 images and is optimized for high-volume workflows.
OpenAI launched GPT Image 2 with the largest ever Image Arena lead of 242 points, beating Google Nano Banana 2. Running on the GPT-5.4 backbone with native reasoning architecture, it achieves 99% character-level accuracy across multiple scripts and supports 4K resolution with batch consistency. Testing across eight categories shows GPT Image 2 winning realism, classical art, lettering, spatial awareness, and image editing, while Nano Banana 2 excels in anime illustration and complex prompt handling. Pricing undercuts Nano Banana 2 at $30 versus $60 per million output tokens.
