Here we are! Issue #3. With a 73% open rate on issue #2 I really need to THANK you all for keeping me motivated! Your readership makes this newsletter better. And to that end: I got some feedback from a reader that suggested I include sample prompts in future newsletters. This one has that! I got feedback from another that a single module could be the whole newsletter. And this week: I’m trying that out.
Less to read, but a lot more detail on one very important update: The new AI image generation model launched to ALL users of ChatGPT (free users included.)
This magazine spread, featuring yours truly, was entirely generated with ChatGPT. Prompt: A two-page editorial magazine spread layout for a business technology magazine. Left page: a large, dramatic full-bleed photograph — a human hand and a robotic arm reaching toward each other across a gap, almost touching, Michelangelo's Creation of Adam style but rendered in a contemporary studio setting. The robotic arm is clearly mechanical but elegant — white and silver. The human hand is brown-skinned and reaches from the left. The background is a deep charcoal gradient. No text on this page. Right page: editorial layout on white. At top, the headline in bold black editorial serif typeface: 'The Year Everyone Finally Understood AI'. Subhead in smaller italic: 'How a new generation of translators turned corporate fear into competitive advantage.' Three-column body text below in clean sans-serif (placeholder but legible). A pull-quote in large type centered in a blue box reads: 'You don't need to understand how the engine works to drive the car.' — Celia Quillian. Bottom right: a small professional headshot of the woman in the image that is attached with the caption 'Celia Quillian, Author of AI for Life and Founder of SmartWorkAI'. The overall layout should look like it could appear in Fast Company or Wired. 16:9 landscape (simulating an open spread).
TRY THIS TODAY / ON MY MIND / HONESTLY THE WHOLE NEWSLETTER
ChatGPT's image update is genuinely wild.
On Tuesday of this week, OpenAI announced the launch of their newest AI Image model Images 2.0. I was one of the lucky few to get early access, so I’ve been playing with it a little under a week now, and I can honestly say it is TRULY impressive. With Google Gemini’s Nano Banana, I thought we were nearing the peak of AI generated quality but…this model really feels like a step up.
So what can it do that others can’t?
First: Any aspect ratio you want (or, up to 8,000 or so.) That means no more generating an image and then taking it into an editing software to crop it down to the right dimension. Previously within ChatGPT, you were limited to square images or those in 2:3 or 3:2 dimensions. This was deeply annoying.
Now you can just request the dimension in your original prompt (9:16, 4:5, panoramic, etc.) and you’re good to go. Relatedly the quality of the image being created is higher (more pixel dense.) That said, the “chat” interface output is not as high quality as when you use the model via API. Doing that unlocks 4k (!!) resolution.
Made by Celia Quillian with ChatGPT. Prompt: Hyper-realistic overhead photograph of a photographer's light table. The table glows from below with cool white light. On it, a 35mm film contact sheet is laid flat — a strip of 36 small film negatives, each a unique photographic frame in the style of 1960s documentary black and white street photography. The frames show scenes: a woman hailing a cab, a child eating an ice cream cone, two men arguing under a street sign, a dog sleeping on a fire escape, a jazz musician performing in a dim bar. A red grease pencil circle marks three of the frames. A loupe (magnifying glass) sits on top of the sheet, partially magnifying one frame. To the left: a yellow legal pad with handwritten notes. Film canister in the corner. 4:5 portrait ratio.
Next: Think and plan, in addition to browsing the web. (On the paid version of ChatGPT, at least.) Let’s say you’re from the smallest of small hometowns. You want to make a “Welcome to Smallville, Alabama” poster. Well, toggle on Web Browsing and Thinking mode, and ChatGPT will scour the web for what it can find about that hometown and use the imagery from its search to inform the planning of the imagery it generates for you. It also means it can make imagery for breaking news developments pretty rapidly… on its own, without necessarily including imagery as context.
Made by Celia Quillian with ChatGPT. Prompt: Make an image of a magazine cover that is illustrated and colorful, almost like a Mod Podge rendering of illustrations that are inspired by Guntersville, Alabama, presently in the modern day. Include landmarks and some well-known locals or historical figures from Guntersville, Alabama. Browse the web before you generate.
And Beyond that: A whole lot more. The level of detail is astounding. Its ability to adhere to stylistic requests is remarkable. It can generate thousands of pictures within a picture. It can generate screenshots depicting multiple open tabs at once down to every detail you request. Infographics. Cartoon panels. “Editing” your OWN photos by generating photos you upload anew (while keeping the things you want to stay the same—like your face—the same.) Fliers with your requested text (or its suggested text!) with an insane level of realism and with few and very far between “AI tells.”
Made by Celia Quillian with ChatGPT. Prompt: A photorealistic cinematic medium shot of a friendly giant — about 40 feet tall, dressed in a rumpled linen shirt and suspenders, with a large brown beard full of small birds nesting in it — seated cross-legged in a flower meadow, carefully holding an ordinary porcelain teacup between two enormous fingers. The teacup is the size of a human torso. Steam rises from it. A full English village is visible behind him — thatched roofs, stone walls, a church steeple — perfectly intact and undisturbed, suggesting he belongs there and everyone has simply adjusted. A villager on a ladder is handing him a biscuit the size of a small table. The giant's expression is delighted and gentle. His eyes are blue-gray and crinkled at the corners. Late afternoon golden hour light. Long shadows across the meadow. The mood is entirely domestic and warm — this is Tuesday afternoon for everyone involved. Shot at a low angle looking slightly up at the giant, keeping the village in soft focus behind. 2.39:1 widescreen.
If you thought we were already in the wild west with AI generated creation… buckle up, friends. It just got wilder. On the one hand, it’s exciting: Anyone can be creative and see their vision come to life if they can provide clear enough instructions. On the other hand, it’s a little scary. Most people still have no idea they are looking at an AI generated image when they see imagery from models that came out two years ago. If you think you are “good” at spotting AI images, get prepared for a wake up call. Because while most will use AI imagery for fun and creativity, others will use it to sell you something you don’t need, or worse: to deceive you and those you love.
This woman does not exist. Catfishers rejoice! Made by Celia Quillian with ChatGPT. Prompt: A vertical iPhone photo, casual and colorful, taken handheld in the bright afternoon. A woman in her late 20s — natural features, a few freckles, hair a little wild from humidity, genuinely happy — stands in front of a wall completely covered in a massive painted mural in the Roma Norte neighborhood of Mexico City. The mural behind her is vivid: hot pink, electric yellow, cobalt blue, figures of women and flowers and birds. She's wearing a burnt orange linen co-ord set, hoop earrings, white Birkenstock sandals. She's holding a michelada in one hand and doing a small shrug-smile at the camera — an "I'm in Mexico and it's noon and I have a drink…sue me~" face. The iPhone's auto HDR has made the mural colors almost unrealistically saturated while her skin looks warm and natural. Shot at roughly eye level, very slightly wide. 9:16 vertical.
So what do we do about it?
If you see something that looks real (especially a news story, or an advertisement that’s promoting astounding outcomes) do yourself (and everyone) a favor and conduct a Google search (or ask ChatGPT or your AI of choice to browse the web!) Find out if the claim is based on reality.
Yeah, yeah, I know. “But I don’t have time for that!” Boy, do I have good news for you! You actually do. Because you’re learning how to get your time BACK with AI in every other area of your life. That’s what you’re subscribed for!
In summary: 1] Know what’s possible. 2] Help OTHERS (i.e. Grandma, your parents, your friends) know what’s possible (like…by forwarding this email!) 3] Triangulate and advocate for the truth. It’s worth the time, for all of our sakes. Trust me.
Image by Celia Quillian with ChatGPT. Prompt: A photograph that appears to have been taken in 2008. Simulate the exact camera characteristics: no flash (nighttime scene), extreme noise grain throughout the image, washed-out pale colors, slight motion blur from a non-steady hand, auto-white balance that has gone slightly purple-green, low dynamic range that crushes both shadows and highlights, and no depth of field (everything is flat-sharp in the harsh way cheap phone cameras render). The scene: a group of five young adults at what appears to be a house party — red Solo cups, someone is mid-laugh, a flat-screen TV in the background shows what looks like Guitar Hero, there is a glass bowl of chips on the coffee table, a person is blurry because they moved. The framing is slightly askew as if taken casually. There is a faint timestamp in the lower-right corner in orange pixel font: '11.22.08 — 11:47 PM'. 4:3 ratio.
Thank you for reading! Please feel free to reply to this email with any feedback on issue #3: what resonated, what didn’t, (did you like the shorter length?) and what questions you have about using AI that you’d love to see my take on. Like what you read? Forward this e-mail to a friend!
Until next time,
Celia
Was this email forwarded to you?
