• AI Toast
  • Posts
  • Google dropped Nano Banana Pro, best image-generation model

Google dropped Nano Banana Pro, best image-generation model

Plus: 5 ways to use Nano Banana Pro

Author

Poonam Soni
November 21, 2025

In partnership with

Welcome to AI Toast!

Here’s what’s on the menu today:

  • Nano Banana Pro

  • Antigravity lands

  • How to use NanoBanana Pro

  • Superhero Tools by Google

  • Quick AI Toasts: Microsoft, NVIDIA, and Anthropic announce partnerships

Total read time: About 7 minutes, perfect for you to fall in love with Google.

If last year’s AI image tools were paintbrushes, Nano Banana Pro is a design studio with a brain. Built on Gemini 3 Pro, it doesn’t just “make pretty pictures”; it reasons, grounds to search, and writes legible text inside images across languages.

  • World-aware visuals: Generate infographics, maps, and educational diagrams that pull in real-world context via search grounding. Think “bike maintenance explainer” or “Newton’s prism” without wobbly facts.

  • Text that doesn’t melt: Clear, correctly rendered copy inside image mockups, posters, UI screens, and even calligraphy. Multilingual too, so localization isn’t a separate workflow.

  • Studio controls: Camera, focus, color grading, lighting shifts (day-to-night), and depth of field. Upscale to 2K/4K and swap aspect ratios without breaking your subject.

  • Consistency window: You can blend up to 14 objects in one image while keeping up to five people consistent. Brand systems, product lines, and cast continuity lead to less unpredictability and more consistency.

  • It’s powerful, not perfect: Small faces and fine details can still slip. Complex blends, masked edits, or dramatic lighting changes can produce artifacts.

Personal Take: Google is clearly positioning image generation as a serious creative workflow, not just a novelty. The jump to 2K/4K, plus camera-level controls, signals a future where AI images move comfortably into design, presentations, education and production work. Try here: Gemini

Startups who switch to Intercom can save up to $12,000/year

Startups who read beehiiv can receive a 90% discount on Intercom's AI-first customer service platform, plus Fin—the #1 AI agent for customer service—free for a full year.

That's like having a full-time human support agent at no cost.

What’s included?

  • 6 Advanced Seats

  • Fin Copilot for free

  • 300 Fin Resolutions per month

Who’s eligible?

Intercom’s program is for high-growth, high-potential companies that are:

  • Up to series A (including A)

  • Currently not an Intercom customer

  • Up to 15 employees

Antigravity is Google’s bet that coding help should evolve from autocomplete to orchestration. Instead of a chatty sidekick, you get agents with a mission control UI, direct tool access, and artifact‑based verification so you can skim outputs, comment, and keep them moving. 

An agentic dev platform that lets AI agents plan, code, run, and verify across editor, terminal, and browser. 

  • Two surfaces: Editor view for hands-on IDE work and Manager surface to orchestrate multiple agents asynchronously. 

  • Trust via Artifacts: Agents produce task lists, plans, screenshots, and recordings for quick review instead of raw logs.

  • Model options: Ships with generous Gemini 3 Pro limits; supports Claude Sonnet 4.5 and GPT‑OSS; free public preview on Mac, Windows, and Linux. 

  • Use cases: End‑to‑end feature builds, UI iteration with browser‑in‑loop verification, long‑running bug hunts, and maintenance in the background. 

  • Live now: Download and explore product docs, changelog, and support from the site.

Personal Take: Early days, but the shape is clear. If agents are going to earn trust in serious codebases, artifacts over logs is the right instinct. 

Superhero Tools (Video Edition)

Gemini: A smart, multimodal AI assistant for chat, writing, coding, and analysis.

NotebookLM: An AI research partner that turns your sources into clear summaries, insights, and audio overviews.

Veo: Google DeepMind’s video generation model for photorealistic, stylized clips from text prompts.

Vertex AI: Google Cloud’s platform to build, tune, deploy, and govern generative AI and agents at scale.

Google Workspace with Gemini: AI woven into Gmail, Docs, Sheets, Slides, and Meet to draft, summarize, analyze, and create.

Google AI Studio: A web IDE to prototype prompts, call Gemini APIs, and ship GenAI apps fast.

NanoBanana Pro in Gemini: Step-by-step tutorial

Google has launched the best model for image generation: Nano Banana Pro. You can use it on Gemini app or website: Try Nano Banana Pro

  • Go to gemini.google.com

  • Inside the Gemini app, select Thinking: In the model dropdown

  • In Gemini’s Create Images 🍌 option to use the latest pro model.

Why Pro: Better text-in-image, higher fidelity, more photos blended, stronger reasoning.

Best prompts (quick templates)

  • Infographic: “Create a clear infographic about Nano Banana; 16:9; clean grid; minimal icons; headline + 3 bullets.”

  • Poster with text: “Design a poster for an AI event happening in Austin, Texas, with bold typography and high contrast; center the title; add a tagline; and ensure the text is sharp and legible.”

  • Style transfer: “Apply the color/texture/style of the reference image to this subject; keep subject identity.”

  • Localized edit: “Turn this scene into nighttime; keep subject lighting natural; add soft bokeh.”

  • Storyboard/grid: “Make a 2x2 storyboard grid, each pane labeled, with a consistent character and cinematic frames.”

Be specific on subject, action, scene, style, aspect ratio, and output quality. Pro understands and renders text reliably.

Use cases (with visual outcomes)

  • Text-heavy designs: Logos, posters, comics, and invite cards with accurate text in multiple languages. 

  • Educational visuals: Infographics and diagrams grounded in real-world knowledge.

  • Advanced edits: Lighting, focus, camera angle, aspect ratio, and color grading.

  • Multi-photo blends: Combine more photos with higher consistency (faces, products, scenes).

Quick AI News Bites

  • The U.S. charges four people for illegally exporting Nvidia GPUs, spurring calls for a Chip Security Act. The scheme allegedly moved 400 A100s via Malaysia; attempts with H100/H200s were blocked. Washington wants location tracking and diversion reporting to curb China-bound high‑end chips.

  • OpenAI rolls out group chats in ChatGPT for up to 20 people across Free, Go, Plus, and Pro. You can invite via link, tag ChatGPT to jump in, and keep personal memory/settings private. It turns ChatGPT into a collaborative space for planning, co‑writing, and research.

  • TikTok adds a Manage Topics setting to tune how much AI‑generated content you see. You can reduce synthetic media (not remove entirely) and draw in more if you like. TikTok is also testing invisible watermarking and funding AI literacy to improve transparency.

  • Prime Video is rolling out AI‑generated video recaps for select originals in the U.S. (beta). They stitch narration, dialogue, and music to summarize key plot points between seasons. A recap button appears on show pages; X‑Ray text recaps remain for spoiler‑safe summaries.

  • Anthropic will scale Claude on Azure and commit up to one gigawatt of compute.NVIDIA and Anthropic form a deep tech partnership to optimize models and future chips. Microsoft ($5B) and NVIDIA ($10B) plan to invest; Foundry customers get access to frontier Claude models.

Boost revenue and gain new customers by partnering with us

Reach over 35K AI enthusiasts with your product.

Join our newsletter to connect with tech professionals, investors, engineers, managers, and business owners worldwide. DM now!

Got feedback, a story worth toasting, or a wild tech question? I’d love to hear from you; just reply, or find me on X.

Cheers,

— Poonam Soni