EP 059 · 71 MIN CHATGPT IMAGE 1.5? STILL NOT CLOSE.
Midjourney v8 Countdown + Nano Banana Pro vs ChatGPT Image 1.5
The short version
Episode 59 has Rory Flynn and Drew Brucker return from the holidays to declare ChatGPT's Image 1.5 update still not close to Nano Banana Pro on realism and color, while Freepik's slow servers push them toward alternatives. Rory demos a Weavy node workflow that turns one Nano Banana Pro render into a full multi-angle shot list with negative-prompting rules, and the pair predict negative prompting will outweigh positive prompting through 2026.
Key takeaways
- 01
ChatGPT's Image 1.5 didn't fix the warm color cast or airbrush-smooth faces they flagged before; Nano Banana Pro remains the clear winner on realism.
- 02
Skip another subscription: build a custom Nano Banana tool in Google AI Studio, connect an API key, and pay only per use.
- 03
On Kling 01, pick a 3–10 second keyframe transition instead of a flat 8 seconds; shorter windows cut from shot to shot cleanly instead of improvising motion.
- 04
Keep the first Nano Banana Pro pass plain and studio-lit, then apply style and branding reference second so detail lands on the subject, not the background.
- 05
In Weavy, split one prompt into parallel nodes with an asterisk so every angle in a shot list generates on its own branch.
- 06
Nano Banana Pro renders real album covers, movie posters and band logos with legible text now, raising new brand and IP risk questions clients didn't have before.
- 07
Rory's prediction for 2026: negative prompting will matter more than positive prompting, because tools like Nano Banana combine an LLM with the image model.
Receipts · said on the record
“We had two things with ChatGPT's image model we would have liked to have seen better. And they didn't fix either one … We still got this 3,300 Kelvin warm tint to every single image, and then we still got the airbrush face.”
“It has the selection between three and 10 seconds … if you're using it on Flow, it's eight seconds for the keyframe to go from start to end, and sometimes that's too much time — it makes stuff up to happen in between the shots.”
“Subjects must never be centered. Rule of thirds, left framing, right framing, negative space, right? Like reject standard portrait framing, you know, cut off the foreheads, put subjects in the corner, like basically just don't do all the normal stuff that makes it look perfect.”
“Negative prompting will be more important than positive prompting in coming future. What you don't want is — my system prompts now have more of what I don't want than what I do want.”
“It's my fault. Don't prompt while sick, guys. Don't do it. Don't post while sick.”
Questions this episode answers
What did Fast Hours think of ChatGPT's Image 1.5 update?
Rory and Drew called it still not close to Nano Banana Pro, saying it kept the warm color cast and airbrush-smooth faces they'd criticized before.
How does Fast Hours use Nano Banana Pro for client car renders?
Rory generates a plain, studio-lit base car in Nano Banana Pro, then layers in a style and branding reference and builds a multi-angle shot list with Weavy nodes.
Why do Rory and Drew like Kling 01 for keyframe video?
Because it lets you pick a 3-10 second transition instead of a fixed 8 seconds, so the model cuts between shots instead of inventing motion in between, unlike Veo 3 on Flow.
What is Rory's prediction about negative prompting?
Rory argues negative prompting will matter more than positive prompting going into 2026, since his own system prompts now list more of what he doesn't want than what he does.
How can you build a custom Nano Banana tool without another subscription?
Rory suggests describing the tool you want in Google AI Studio, connecting your own API key, and paying per use instead of a subscription markup.
Show notes
Drew and Rory stumble back from the holiday chaos—one fresh off vacation, the other barely resurrected from a mystery NYC illness.
Between fever dreams and booger fingers, they somehow manage to tear into ChatGPT's Image 1.5 disappointment, expose why Nano Banana Pro is quietly dominating their workflows, and reveal the Weavy automation setup that's actually working (while FreePik continues its reign of mediocre terror).
The duo gets brutally honest about why OpenAI feels like it's slipping, why negative prompting might be more important than what you actually want to create, and how to build your own custom AI tools in Google AI Studio without selling your soul to another subscription.
Plus: vintage Kodak rally cars, the art of perfect thumbnails, coconut water in cocktails, and why their illness prevention protocols involve more vitamin C than common sense.
If you survived their holiday absence and made it through the mandatory 20-minute ramble tax, you'll be rewarded with legitimate workflow gold that actually ships.
