EP 035 · 90 MIN IT WROTE A FULL PAGE OF TEXT. FLAWLESSLY.
ChatGPT Image Gen Just Went Nuclear! + A (Final) Midjourney V7 Preview
The short version
Episode 35 reacts to ChatGPT's new native image generator landing days before Midjourney's long-awaited V7. Rory and Drew are most struck by its accurate full-paragraph text rendering and conversational, in-context editing, and share workflows for wireframe-to-ad-layout templating, JSON style-profile extraction and Figma-annotated inpainting. They also hands-on test Reve and Ideogram against Midjourney, dissect the Studio Ghibli meme wave's ethics and virality, and preview V7's office-hours notes, including a Smart Select demo and Midjourney's intentional image underexposure.
Key takeaways
- 01
Feed ChatGPT a wireframe as an image reference instead of describing a layout in text: Rory's Figma-built ad template got filled in almost perfectly on the first try.
- 02
When ChatGPT nails an output, ask it to return the exact prompt it used: that prompt then reliably transfers into Midjourney or Reve too.
- 03
To edit just part of an image in ChatGPT, annotate it first (circle or highlight the area in red in Figma) rather than describing the region in words.
- 04
ChatGPT can extract a reusable JSON style profile from a batch of reference images — a fast way to reverse-engineer someone else's aesthetic.
- 05
Midjourney's images run about one to two stops underexposed on purpose, per Drew's photography background: it protects highlight detail and leaves more room to grade in post.
- 06
Ideogram's outputs are too coherent and 'sterile' for Rory's taste — Midjourney's unpredictability is the point, not a flaw to fix.
- 07
A new image tool has to be either hyper-niche or dramatically better across the board to break an existing workflow: a couple of cool tricks isn't enough.
Receipts · said on the record
“That's goddamn wild because it was all text. And that was the biggest surprise was how I was nailing the text. … That's like whole page worth of writing of text. And I'm like, that's seriously impressive.”
“Can we talk about the best outcome scenario for that dude ever? He says AI is trash, and then everyone seemingly trolls him all over the internet by making things in his style.”
“If I'm doing some photography … I always underexpose a little bit. And the reason for that is like, sometimes you might catch a light that then can be blown out. It's called like a blown out exposure and you can't fix that in post.”
“It's either going to be super niche or it's got to be the king. You know what I mean at this point? And it can't really be anything in the middle.”
“That was my feedback. There's no risk. It doesn't take any risk. Mid-journey is all risk. That's why it's great. Mid-journey is full send on every prompt. You never know what you're going to get, which is great.”
Chapters
- 0:00:00 Catching Up: Beer, Pasta & Big AI Moves
- 0:03:08 Image Gen Wakes Up After a 6-Month Nap
- 0:05:59 ChatGPT Goes Nuclear (And We Have Thoughts)
- 0:08:50 How Creators Are Actually Using These Tools
- 0:11:45 The Language of Prompting Is Getting Weird
- 0:14:55 AI Image Gen Tools Are Taking Different Paths
- 0:17:56 Creative Flow in the Age of Automation
- 0:21:05 Ethics, Aesthetics, and a Bit of Chaos
- 0:33:32 What Makes Midjourney Still Feel Magical
- 0:36:09 Prompt Tips That Actually Work
- 0:38:53 How Templates + Moodboards Shape Results
- 0:41:34 The Community Is the Underrated Cheat Code
- 0:44:51 Fixing a Janky Workflow, One Prompt at a Time
- 0:48:37 Let's Talk Upscaling: The Good, Bad & Blurry
- 0:55:59 Where AI Workflow Tools Need to Go
- 0:59:59 Midjourney V7: The Rumors, the Hope, the Hype
- 1:04:06 AI Upscaling Is Getting... Surprisingly Good
- 1:11:22 Why the AI Dev World Feels Different Now
- 1:17:25 Who’s Winning the AI Tool War (For Now)
- 1:23:58 Wild New Features We Didn’t See Coming
- 1:28:59 Wrap-Up
Questions this episode answers
What's new about the ChatGPT image generation discussed in this episode?
It's OpenAI's native image generator, shown on-screen as "4o Image" and replacing the old DALL-E integration. Rory and Drew highlight its accurate long-paragraph text rendering, conversational in-context edits ("cut the saturation, add contrast"), and the ability to follow a wireframe or annotated reference image far more precisely than a text description alone.
How does Midjourney's Omni-Reference factor into this episode?
It hadn't shipped yet. Rory says that if Omni-Reference lands well when V7 arrives, it will be the pillar feature that matters most to users, even more than aesthetic upgrades, because reliable, editable character consistency is what most people actually want.
What did Fast Hours think of Reve?
Rory's hands-on testing found Reve strong on fashion and luxury photorealism, with convincing skin blemishes and imperfect hair, but weaker on more specific or complex scenes like cars or licensed characters. Drew argues a challenger tool like Reve needs to be either hyper-niche or dramatically better everywhere to actually change anyone's workflow.
Why do Rory and Drew say Midjourney images look underexposed?
Drew, drawing on his photography background, explains that Midjourney intentionally renders roughly one to two stops darker than a 'correct' exposure, which protects highlight detail from being unrecoverably blown out and leaves more room to grade the image afterward.
What's the Studio Ghibli meme trend the hosts discuss?
A viral wave of turning photos and film stills into Studio Ghibli-style images using ChatGPT's new generator, after a co-founder of the studio had criticized AI art. Rory and Drew call it the best unintentional PR campaign imaginable, while also flagging real risks, like someone using the same tooling to mock up a fake gift-card scam image.
Show notes
Fast Hours Podcast—Ep. 35: ChatGPT Image Gen Went Nuclear! + The Final Countdown to Midjourney V7 Preview (with Drew Brucker & Rory Flynn)
ChatGPT’s new image generator (not named Dall-E?) is here, and Midjourney V7 is just around the corner.
In this episode, Drew Brucker and Rory Flynn dive into the latest advancements in AI image generation, from ChatGPT’s surprisingly powerful new image tool to a look at what’s coming in Midjourney V7. They explore how these tools are reshaping creative workflows, what features matter most, and where things might be headed next.
You’ll also hear about: ChatGPT Image gen use cases and functionalities Whether ChatGPT is a serious visual competitor to Midjourney How AI tools like Reve are shifting the creative landscape The viral nature of AI visuals in social media The ethics, weirdness, and opportunities of AI in creativity
Oh, and there’s a quick recap of their first-ever IRL meeting in New York.
⏱️ Midjourney Fast Hour [00:00] Catching Up: Beer, Pasta & Big AI Moves [03:08] Image Gen Wakes Up After a 6-Month Nap [05:59] ChatGPT Goes Nuclear (And We Have Thoughts) [08:50] How Creators Are Actually Using These Tools [11:45] The Language of Prompting Is Getting Weird [14:55] AI Image Gen Tools Are Taking Different Paths [17:56] Creative Flow in the Age of Automation [21:05] Ethics, Aesthetics, and a Bit of Chaos [33:32] What Makes Midjourney Still Feel Magical [36:09] Prompt Tips That Actually Work [38:53] How Templates + Moodboards Shape Results [41:34] The Community Is the Underrated Cheat Code [44:51] Fixing a Janky Workflow, One Prompt at a Time [48:37] Let's Talk Upscaling: The Good, Bad & Blurry [55:59] Where AI Workflow Tools Need to Go [59:59] Midjourney V7: The Rumors, the Hope, the Hype [01:04:06] AI Upscaling Is Getting... Surprisingly Good [01:11:22] Why the AI Dev World Feels Different Now [01:17:25] Who’s Winning the AI Tool War (For Now) [01:23:58] Wild New Features We Didn’t See Coming [01:28:59] Wrap-Up
👉 Ideal for: creative professionals, marketers, designers, and AI art nerds 🔔 Subscribe for weekly episodes on Midjourney, AI tools, and next-gen creativity. 🎧 Listen now on YouTube, Spotify, or Apple Podcasts.
