Uploaded November 2025 | Updated November 2026, 3 days ago
Google and DeepMind have just released Nano Banana Pro, the highly anticipated image model powered by Gemini 3. I got early access to test it out, and have exclusive insights you won't find anywhere else.
In this video, I push Nano Banana Pro to its limits. We explore where it shines, where it's still a little "mushy," and how it handles complex tasks. We dive into the new features, including advanced aspect ratios, outpainting, and incredible style transfer capabilities (wait until you see the Zelda example!).
The Real Game Changers? 3D Space and Text. The biggest upgrades are the model's surprising understanding of 3D space—allowing us to create virtual sets and correctly break the 180 rule—and its phenomenal text handling. We even generated a full comic book page just from the script!
Plus an exclusive Interview with DeepMind Don't miss our exclusive chat with Nicole Brichtova, Product Lead for Image and Video at DeepMind. We discuss the future of AI interfaces, when consistency will be solved, how Nano Banana integrates with Gemini 3 and V3, and the ultimate "North Star" for the project.
Big Thanks to today's Sponsor Vibecode! vibecode.go.link/4niCg
Get your first 3 apps free using code “Media”!
LINKS:
NANOBANANA In Gemini 3: gemini.google.com
Google Deepmind: https://deepmind.google/
Henchmen Inc Comic: theoreticallymedia.gumroad.com/l/henchmen
Chapters:
00:00 - Intro: Nano Banana Pro is Here!
00:34 - Gemini 3: The "Thinking" Image Model
01:13 - Where to Access It (AI Studio, API, Flow)
01:20 - Benchmark Test: Wine Glass and Clock
01:38 - Pelican Bike SVG Test
01:45 - The ULTIMATE Test (Pelican, Bike, Wine, VR!)
01:58 - New Feature: Dynamic Aspect Ratios
02:12 - Outpainting Test (Flamethrower Girl)
02:32 - Thoughts on Straight Image Generation Quality
02:47 - Key Strength: Style Transfer (Zelda Example)
03:10 - Weakness: Struggling with Exact Replication
03:26 - The "Context Jumbling" Issue in Gemini
03:50 - When Context Jumbling Creates ART
04:11 - Where Nano Banana Pro Flies: Spatial Awareness
04:22 - Creating Virtual Sets
04:43 - Populating Scenes with Characters
05:05 - The Biggest Win: Moving the Virtual Camera
05:26 - Breaking the 180 Rule Correctly
05:47 - Solving the "Paper Doll" Flip Issue
05:58 - Mind-Blowing Text Handling
06:10 - Generating a Comic Book From Script
06:35 - Comparing AI Comics vs. The Real Page
07:04 - AI Adding Its Own Context (Joker Letter)
07:23 - Tips for Multi-Image Referencing
07:47 - Overall Early Impressions (Pros and Cons)
08:27 - Sponsor Segment
11:17 - Interview: Nicole Tovar (DeepMind Product Lead)
11:42 - The "Thinking Model" Update Surprises
12:34 - Does the Model Understand 3D Space?
13:20 - The DeepMind Ecosystem (Nano Banana, V3, Gemini 3)
13:47 - DeepMind's Approach to Integrating Models
14:40 - Long-Term Vision for AI Modalities
15:19 - The "Ketchup" Naming Mystery Solved
15:37 - The Future of the User Interface
15:58 - Will UIs Become Real-Time and Spatial?
16:42 - Definitely Not Nodes
17:03 - When Will Consistency Be Cracked?
17:39 - The "North Star" of Nano Banana
18:32 - Outro and Final Thoughts
👇 What feature are you most excited to try in Nano Banana Pro? Let us know in the comments!
If you found this deep dive helpful, please LIKE the video and SUBSCRIBE for more updates on the rapidly evolving world of AI!
Google and DeepMind have just released Nano Banana Pro, the highly anticipated image model powered by Gemini 3. I got early access to test it out, and have exclusive insights you won't find anywhere else.
In this video, I push Nano Banana Pro to its limits. We explore where it shines, where it's still a little "mushy," and how it handles complex tasks. We dive into the new features, including advanced aspect ratios, outpainting, and incredible style transfer capabilities (wait until you see the Zelda example!).
The Real Game Changers? 3D Space and Text. The biggest upgrades are the model's surprising understanding of 3D space—allowing us to create virtual sets and correctly break the 180 rule—and its phenomenal text handling. We even generated a full comic book page just from the script!
Plus an exclusive Interview with DeepMind Don't miss our exclusive chat with Nicole Brichtova, Product Lead for Image and Video at DeepMind. We discuss the future of AI interfaces, when consistency will be solved, how Nano Banana integrates with Gemini 3 and V3, and the ultimate "North Star" for the project.
Big Thanks to today's Sponsor Vibecode! vibecode.go.link/4niCg
Get your first 3 apps free using code “Media”!
LINKS:
NANOBANANA In Gemini 3: gemini.google.com
Google Deepmind: https://deepmind.google/
Henchmen Inc Comic: theoreticallymedia.gumroad.com/l/henchmen
Chapters:
00:00 - Intro: Nano Banana Pro is Here!
00:34 - Gemini 3: The "Thinking" Image Model
01:13 - Where to Access It (AI Studio, API, Flow)
01:20 - Benchmark Test: Wine Glass and Clock
01:38 - Pelican Bike SVG Test
01:45 - The ULTIMATE Test (Pelican, Bike, Wine, VR!)
01:58 - New Feature: Dynamic Aspect Ratios
02:12 - Outpainting Test (Flamethrower Girl)
02:32 - Thoughts on Straight Image Generation Quality
02:47 - Key Strength: Style Transfer (Zelda Example)
03:10 - Weakness: Struggling with Exact Replication
03:26 - The "Context Jumbling" Issue in Gemini
03:50 - When Context Jumbling Creates ART
04:11 - Where Nano Banana Pro Flies: Spatial Awareness
04:22 - Creating Virtual Sets
04:43 - Populating Scenes with Characters
05:05 - The Biggest Win: Moving the Virtual Camera
05:26 - Breaking the 180 Rule Correctly
05:47 - Solving the "Paper Doll" Flip Issue
05:58 - Mind-Blowing Text Handling
06:10 - Generating a Comic Book From Script
06:35 - Comparing AI Comics vs. The Real Page
07:04 - AI Adding Its Own Context (Joker Letter)
07:23 - Tips for Multi-Image Referencing
07:47 - Overall Early Impressions (Pros and Cons)
08:27 - Sponsor Segment
11:17 - Interview: Nicole Tovar (DeepMind Product Lead)
11:42 - The "Thinking Model" Update Surprises
12:34 - Does the Model Understand 3D Space?
13:20 - The DeepMind Ecosystem (Nano Banana, V3, Gemini 3)
13:47 - DeepMind's Approach to Integrating Models
14:40 - Long-Term Vision for AI Modalities
15:19 - The "Ketchup" Naming Mystery Solved
15:37 - The Future of the User Interface
15:58 - Will UIs Become Real-Time and Spatial?
16:42 - Definitely Not Nodes
17:03 - When Will Consistency Be Cracked?
17:39 - The "North Star" of Nano Banana
18:32 - Outro and Final Thoughts
👇 What feature are you most excited to try in Nano Banana Pro? Let us know in the comments!
If you found this deep dive helpful, please LIKE the video and SUBSCRIBE for more updates on the rapidly evolving world of AI!










