PM at Google DeepMind working on Gemini's image gen capabilities. In my 2 era! Opinions are my own.
San Francisco, CA
Joined December 2015
- Tweets483
- Following196
- Followers4.8K
- Likes596
Pinned Tweet
With Nano Banana 2 is the first model to ground in images from web search! I took it for a spin last night and wanted to share some of the beauty of Pakistan through a series of watercolors!
My parents' favourite bridge
Prompt: "Generate a watercolor of the Lansdowne and Ayub Bridges"
Thats 1.35 billion images on average in just Gemini App since that bananas day that was August 26, 2025!
More than 50 billion images have been generated with our Nano Banana models. 🍌
#GoogleIO
ALT A transparent glass cube overflowing with yellow bananas sits on a reflective surface against a purple background, overlaid with the text "Nano Banana 50B+ images generated."
Naina Raisinghani retweeted
Creative teams stretch the limits of Nano Banana Pro. Check out our work with Studio Porto Rocha 👉 yoyoyo.ai/
Beyond generating high-fidelity visuals, we wanted to test the limits of what Nano Banana Pro can do. We worked with design partners Porto Rocha to build out a hypothetical brand called YOYOYO to see how the model would handle the task. Here’s what we found:
🎨Brand consistency: Across logos, colors, and typography, the model maintained a strict, cohesive brand identity (even for wildly diverse concepts)
🛍️Environmental realism: We asked to see the products in storefront and studio mockups. It nailed accurate lighting, shadows, and physical proportions - even when upscaled for massive retail displays
🪀Spatial accuracy: We tested spatial volumes for physical packaging. The generated proportions were so precise that we were able to 3D-print the functional yo-yo
How have you been pushing the limits of Nano Banana Pro? Let us know in the replies below!
Yoyoyo!
What happens when designers get ahold of Nano Banana Pro?
The results are beautiful compositions that I haven't seen before. It was also a great learningexperience in how our users use the model.
Learn more at yoyoyo.ai
kudos also to @heyitsbeaaaaaaa
Beyond generating high-fidelity visuals, we wanted to test the limits of what Nano Banana Pro can do. We worked with design partners Porto Rocha to build out a hypothetical brand called YOYOYO to see how the model would handle the task. Here’s what we found:
🎨Brand consistency: Across logos, colors, and typography, the model maintained a strict, cohesive brand identity (even for wildly diverse concepts)
🛍️Environmental realism: We asked to see the products in storefront and studio mockups. It nailed accurate lighting, shadows, and physical proportions - even when upscaled for massive retail displays
🪀Spatial accuracy: We tested spatial volumes for physical packaging. The generated proportions were so precise that we were able to 3D-print the functional yo-yo
How have you been pushing the limits of Nano Banana Pro? Let us know in the replies below!
No trip back home to Karachi is complete without going to Dolmen Mall. Ask any Karachiite!
Prompt: "Generate a water color of Dolmen Mall Karachi"
Where do you want Nano Banana 2 to take you?
@GeminiApp, @Google AI Mode, @GoogleAIStudio, @FlowbyGoogle and Vertex!
Nearly 1 year ago, we were the first to launch a model that not only generated images but also supported conversational, multi-turn editing. Today we bring you our best model yet - Nano Banana 2! 🍌🍌 When working on this model we had one goal - Pro level capabilities at Flash level costs! And we have delivered the potassium!
blog.google/innovation-and-a…
Naina Raisinghani retweeted
Interactive media is here. Project Genie🧞is live for Ultra subscribers in the US.
Favorite use case - turning giant stuffed animals into playable characters. Meet the 🍌🦆.
What object on your desk would you turn into a character first?
Looking into 2026, what a time to be alive (and online). 💻 Being an AI consumer in this era is a genuine privilege. From LLMs to creative tools, the sheer variety is mind-blowing. Happy back to work to all! 🥂👩🏾💻
Had to confirm with Kaushik offline that this was indeed one prompt. 🤯
A fun test to see how Nano Banana Pro 🍌 can handle very complex prompts (getting the heights/labels of the bars/pie slices/arrows all correct). We've come a long way in 2025. (Prompt in thread).
ALT Generate an image of a bar chart with 10 items (labeled with the last 10 letters of the alphabet) on the x axis, each with unique values ranging from 1 to 100 (also labeled on top of the bars). On the right side should be a pie chart with 60 for A (red), 25 for B (blue), and 15 for C (green). On the bottom right, show me a mini infographic with labeled major components of a car.
Me today -> Me OOO tomorrow!
Here's a prompt to make cute little diorama lamps:
Prompt: A [subject] is [doing something] in a [place]. Isometric 3D cube diorama with internal lighting, cute chibi figurine style, matte PVC material, big head/small body. Neutral background on a table in a dark room. Enclosed in a cube. Fill the diorama with many cute little details. Add delicate imperfections like dust, scratches and textures to make it real.
Things you ALL asked for:
Improvements:
- Improved instruction following esp. over many turns
- Character consistency esp. for Asian faces + kids.
- Improved adherence to references
New capabilities:
- Transparency
- Masked based editing/ editing only specific parts of the image, while leaving the rest unchanged.
- Adherence to brand guidelines
👋🏽 I'm a PM on Nano Banana, let's just call me the Naina that goes banana over images. If you have any feature requests you find a-peel-ing or any big ideas you want to plantain into the team's brains, comment away!
cc @nbrichtova @heyitsbeaaaaaaa @oliver_wang2 @m__dehghani @19kaushiks
👋🏽 I'm a PM on Nano Banana, let's just call me the Naina that goes banana over images. If you have any feature requests you find a-peel-ing or any big ideas you want to plantain into the team's brains, comment away!
cc @nbrichtova @heyitsbeaaaaaaa @oliver_wang2 @m__dehghani @19kaushiks