I’m always on the lookout for the next big thing in AI, but I was not ready for this. I just saw a video from an AI professional testing Google’s new image model, Gemini 2.5 Flash, also known as “Nano Banana”, and it’s an absolute game-changer for image editing and generation.
The YouTuber kicked things off with a wild example. He took a thumbnail of MKBHD holding two phones and simply prompted the AI, “Flip the phones over.” The result was stunning! The AI knew exactly what the back of an iPhone looked like and even rendered the correct operating systems on the flipped screens. I was blown away.
Here’s a breakdown of the incredible things this innovator demonstrated:
✨ Mind-Blowing 3D & Character Consistency
This is where it gets crazy. The model has a stunning grasp of 3D space and keeping characters consistent.
- 3D Rotation: The expert prompted it to show a can of Coke from three different angles, and the result was flawless. It even spun characters around 180 degrees, accurately predicting what their backs would look like.
- Adding Objects: The creator uploaded a photo of himself and had the AI add a can of Coke for him to hold. Then, he had it add reflective glasses, and you could actually see the background reflected in the lenses. So cool!
- Character Sheets: He turned a photo of Sam Altman into an anime character, then a 3D model, and finally generated a full character sheet with different poses and angles. This is huge for artists and creators.
⚙️ Unbelievable In-Image Editing
Editing existing photos is where this tool really shines. I’ve never seen anything this seamless.
- Zooming Out: The YouTuber took the classic moon landing photo and asked the AI to “zoom out and show me all of this on a sound stage.” The AI built an entire Hollywood set around the original image while keeping the lighting and grainy style perfectly consistent.
- Object Removal: The creator took a photo of the OpenAI founders and flawlessly removed people one by one. You would never know they were there.
- Adding People: In that same photo, the expert added a picture of himself, and the AI blended him in perfectly, matching the lighting, shadows, and even his facial expression to the scene.
🚀 How To Try It Yourself
The person who shared it also showed two easy ways you can access this model right now. It’s currently free but might be rate-limited.
- Google AI Studio: Go to studio.google.com and select “Gemini 2.5 flash image preview” under the featured models.
- Gemini: In the regular Gemini chat, select the “2.5 flash” model. Then, click the three dots in the bottom corner and turn on “Image generation.”
💡 Other Awesome Capabilities
This AI professional packed the video with even more tests:
- 📌 Photo Restoration: It took a damaged, black-and-white photo, repaired all the cracks, and colorized it with incredible accuracy.
- 📌 Text Editing: It changed the text in a thumbnail from “BIG NEWS” to “HUGE NEWS” while keeping the exact font and style.
- 📌 Physics & Lighting: The model generated a photo of a car with a person’s reflection perfectly visible in the headlights and grill. The detail is just phenomenal.
Seriously, the character consistency and its understanding of physics and light are unlike anything I’ve seen before. It’s leading the pack on the image editing leaderboards for a reason.
For the full deep-dive and to see all of these mind-bending examples for yourself, make sure to watch the original video from the creator!