The short answer
Pick Nano Banana for image edits that need to keep the same person or product, like e-commerce shots or a costume on a real photo. Try ChatGPT's Images 2.0 if text inside the image matters to you. Sam liked it for that; Jordan was underwhelmed.
Our evidence here is thin. We built three things with Nano Banana, but we only spent a few days with Images 2.0 before discussing it in Episode 40 (April 2026). Treat this as a first impression, not a test.
What we built with each
Nano Banana
- AI product images for a 3D-printed hoodie charm store: Sam photographed his daughter's printed charms, dropped them into Google AI Studio and asked Nano Banana to put them on a yellow hoodie, then on a model (Episode 12, at 13:27).
- Trick-or-treat AI costume builder: Jordan turned webcam photos into costumed images, calling Nano Banana through Replicate from a Replit app, then animated them with Google Veo (Episode 18).
- HipTrain Meta ad concepts in Claude Design: Claude Design wrote the image prompt, Sam ran it in Nano Banana, and fed the image back into the ad (Episode 40, at 05:12).
Jordan also uses it for the generated backgrounds he makes for almost every episode, and showed it off in AI Studio with a clay figure of a black lab eating a hot dog (Episode 38).
ChatGPT images
- Ryz Labs Autoblogger: OpenAI's image models illustrate the blog posts, with FLUX added for variety so the images do not all look alike (Episode 2).
- Jordan's Episode 40 background, made with Images 2.0.
Head-to-head
Why continuity matters: when Nano Banana launched in August 2025, Sam asked what the hype was about. Jordan's answer was that it can merge two images, like a person and a dog placed together on a basketball court, and edit a scene while keeping the subject the same. He pointed to interior design (add a couch or a wall to a room) and e-commerce product shots as the obvious uses (Episode 10, at 08:59). That is exactly what Sam did weeks later, prompting the charm images onto a girl in black pants and white shoes without losing the charms.
The only time we compared them directly was Jordan swapping his usual background tool:
Yeah. I've been playing with it a little bit. I still I thought I was a little underimpressed, to be honest. I think some of the examples they showed were more impressive than the results I was getting.
— Jordan Metzner, Episode 40
He added at 15:13 that he did not know if there was "a significant difference" from the backgrounds Nano Banana had been making. Sam was more positive. He said older models would turn a name like Jordan into gibberish, and that Images 2.0's text fidelity was the exciting part.
Where each one fell short
ChatGPT. For us the issue was continuity. When Jordan built his AI image editing chatbot in Episode 10, he noted OpenAI had image generation "for a while, but not continuity." In the Autoblogger, the worry was sameness, which is why FLUX was added. And Images 2.0 did not beat its own demo examples for Jordan.
Nano Banana. Likeness. In the costume builder Jordan said "I don't know if those are exactly my eyes," though his body and office came through. Nothing on the show suggests it struggles with text, but we never tested it.
For a third opinion: in Episode 50, Sam tried Meta's new Muse Image and said it "feels pretty similar to Nano Banana or OpenAI"; see Meta Muse Image vs Nano Banana. No guest compared ChatGPT and Nano Banana on the show.
Which one we reach for now
Nano Banana. Sam's verdict from the charm store still sums it up:
Every single time it's come out perfect, then I prompt it again.
— Sam Nadler, Episode 12
In December 2025 Jordan called Nano Banana "the best image generation as far as the largest models on the market" (Episode 23). By April 2026 he saw ChatGPT getting "deeper" into images to compete where Google "had almost, like, run away with it." His view is that more models and cheaper images are good for everyone. If text in the image is your priority, Images 2.0 is worth a test; for edits that must keep a person or product consistent, we would start with Nano Banana. For the wider model comparison, see Gemini vs ChatGPT.
FAQ
Is Nano Banana better than ChatGPT for images?
For our uses, yes so far. Sam's Nano Banana product shots for his daughter's charm store were right on the first prompt almost every time, and when Jordan switched his show background to ChatGPT Images 2.0 he could not see a significant difference.
Is ChatGPT Images 2.0 good at text in images?
That was Sam's main takeaway in April 2026: reports said its text fidelity is really good, where older models turned a name like Jordan into gibberish. We did not run a formal text test.
What is Nano Banana best at?
Continuity: editing or merging images while keeping the subject the same. We used it for e-commerce product shots, costume images from webcam photos and ad creative.
Where can I use Nano Banana?
We run it inside Google AI Studio, and Jordan called it through Replicate for the trick-or-treat costume builder.