# Nano Banana Pro Published articles for Nano Banana Pro. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## WorldCast: Exploring Intelligent Visual Generation in World State Prediction DevFeed: [WorldCast: Exploring Intelligent Visual Generation in World State Prediction]() Original publisher: [Read original article]() Author: Wenhao Yan, Haodong Li, Xiaotong Li, Yaxin Luo, Zhengrong Yue, Jialv Zou, Xu Huang, Haojin Yang, Nan Jia, Yan Feng, Peng Pei, Xunliang Cai Published: 2026-10-23T04:00:00Z Content type: article Language: en Sources: [arXiv: Artificial Intelligence]() Topics: [Computer vision](), [Artificial Intelligence](), [World models](), [Image](), [stable-diffusion]() Tags: [benchmark](), [capabilities](), [cs-ai](), [cs-cv](), [forecast](), [future](), [generation](), [generative-models](), [gpt](), [gpt-image](), [knowledge](), [language-models](), [multimodal](), [nano-banana-pro](), [open-source](), [performance](), [planning](), [real-world](), [reasoning](), [state](), [world]() ### AI overview WorldCast evaluates multi-step visual world-state prediction: forecasting successive future states from multimodal observations and references, combining understanding, generation, and reasoning. Its 320 manually designed cases span eight categories, and the study evaluates 12 models. GPT-Image-2.5 System scores 72.8, which the authors describe as early-stage prediction ability on real-world tasks; Nano Banana Pro scores 42.8, while the best open-source model scores 20.1. Reported weaknesses include complex context, entangled references, and reasoning-heavy tasks. The authors also find that strong generative models draw effectively on their underlying language models. These results indicate substantial capability gaps and point to dynamic planning, reasoning, and unified multimodal modeling as remaining challenges. ### Source excerpt arXiv:2610.20731v2 Announce Type: replace-cross Abstract: Humans interpret multimodal observations of the world and forecast how it will evolve. Recent multimodal generative systems have demonstrated similar capabilities: they can integrate multimodal context and synthesize visual outputs that depict changes in the world. To characterize this emerging trajectory, we organize model capabilities into a four-level hierarchy and discover that higher-level tasks remain relatively underexplored despite model progress. To bridge this gap, we explore model performance on multi-step world state prediction, a high-level task that requires models to forecast successive future world states from multimodal observations and diverse references. Such a task requires models to possess entangled understanding and generation capabilities, as well as extensive world knowledge to translate human action sequences, physical transformations, and abstract visual reasoning into coherent visual outputs. We introduce WorldCast, a benchmark comprising 320 manually designed cases across eight categories, and evaluate 12 representative models. The conducted experiments yield three key insights: (a) advances in frontier models are rapidly pushing the boundaries of intelligent world modeling, with the GPT-Image-2.5 System scoring 72.8 and demonstrating early-stage visual state prediction capabilities on real-world tasks; (b) WorldCast reveals substantial gaps between models: even Nano Banana Pro reaches only 42.8 and the best open-source model scores 20.1, with deficits in handling complex context, entangled references, and reasoning-heavy tasks; (c) strong generative models effectively draw on the capabilities of their underlying language models. These findings highlight the need for multimodal generative models that are capable of dynamic planning and reasoning while also being deeply unified to faithfully simulate real-world states. ## Generate your own Cinematic Video Overviews in NotebookLM. DevFeed: [Generate your own Cinematic Video Overviews in NotebookLM.]() Original publisher: [Read original article]() Author: Pete Aykroyd Published: 2026-03-04T17:00:00Z Content type: release Language: en Sources: [Gemini Notebook]() Topics: [Generative AI](), [generative-ai-tools]() Tags: [ai](), [gemini](), [google](), [google-ai](), [google-ai-ultra](), [google-labs](), [nano-banana-pro](), [none](), [notebooklm](), [video]() ### AI overview NotebookLM is launching Cinematic Video Overviews, which use AI models including Gemini 3, Nano Banana Pro, and Veo 3 to create tailored videos with animations and detailed visuals. Gemini selects the narrative, visual style, and format, then refines its work for consistency. The feature is available in English to Google AI Ultra subscribers aged 18 and older on web and mobile. ### Source excerpt NotebookLM is introducing Cinematic Video Overviews, a major update to its AI-powered video creation capabilities. This new feature moves beyond narrated slides in Video... ## Build with Nano Banana Pro, our Gemini 3 Pro Image model DevFeed: [Build with Nano Banana Pro, our Gemini 3 Pro Image model]() Original publisher: [Read original article]() Author: Alisa Fortin Published: 2025-11-20T15:11:14Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [ai-image-generator](), [Image](), [multimodal-ai](), [image editing]() Tags: [4k](), [adobe](), [agentic](), [agents](), [ai](), [ai-benchmarks](), [ai-generated-media](), [ai-studio](), [antigravity](), [api](), [applications](), [apps](), [artistic-style](), [assets](), [canvas](), [gemini](), [image](), [image-generation](), [model](), [nano-banana-pro](), [none]() ### AI overview Nano Banana Pro (Gemini 3 Pro Image) is a higher-fidelity image generation and editing model, available in paid preview through the Gemini API in Google AI Studio and Vertex AI. It offers controls over lighting, camera, focus, color grading, and composition; resolutions up to 2K and 4K; improved text rendering and image localization; and the ability to combine reference images, including products and logos. Google Search grounding can provide real-time information for data-driven images when enabled. SynthID watermarks are integrated into every image it creates or edits. The excerpt positions Gemini 2.5 Flash Image as faster and lower cost, while Gemini 3 Pro Image offers higher quality at higher cost and latency. ### Source excerpt Nano Banana Pro, or Gemini 3 Pro Image, is our most advanced image generation and editing model. ## Introducing Nano Banana Pro DevFeed: [Introducing Nano Banana Pro]() Original publisher: [Read original article]() Author: Naina Raisinghani Published: 2025-11-20T15:05:02Z Content type: release Language: en Sources: [Google DeepMind News]() Topics: [ai-image-generator](), [Image](), [image editing](), [synthid](), [generative-ai-tools](), [Color Grading]() Tags: [deepmind](), [generation](), [google](), [image](), [image-generation](), [model](), [nano-banana-pro](), [none]() ### AI overview Nano Banana Pro (Gemini 3 Pro Image) is Google's image generation and editing model for complex, high-quality compositions; the article contrasts it with the original Nano Banana for faster, simpler editing. It uses Gemini 3 reasoning and real-world knowledge, with Google Search access for some current-information visuals. Features include legible in-image text in multiple languages, blending up to 14 images while preserving the resemblance of up to five people, localized editing, camera and lighting controls, and 2K or 4K output. Access is rolling out across Gemini, Search, NotebookLM, Ads, Workspace, developer tools, Vertex AI, and Flow, with availability and quotas varying by product and subscription. Google says generated media carries an imperceptible SynthID watermark; Gemini users can upload images and ask whether Google AI generated them, initially using English prompts. Visible Gemini sparkle marks remain for free and AI Pro users, while Google AI Ultra and AI Studio outputs omit them. ### Source excerpt Nano Banana Pro is our new image generation and editing model from Google DeepMind.