The Official ImageGen, 4o and Dall-E Megathread

…with 4o…

Prompt

Ultra-realistic wide studio beauty macro portrait featuring two 20-something skeleton models standing side-by-side:

• On the left: a woman with fair skin, faint freckles, slightly glossy lips, and well-groomed features, wearing a traditional red hanbok with floral embroidery.

• On the right: a man with fair skin, faint freckles, clean-shaven, wearing a white bathrobe with visible texture.

Both models are skeletal but retain realistic skin texture, visible brow hairs, and partial facial features (eyes, brows, hair) above the jawline. Their full heads and hair are entirely visible. Each holds a soft-pink minimalist skincare tube clearly displaying:

• Top: a realistic dung beetle rolling a ball (logo-style graphic)

• Center line 1 (brand name): bold text “Dung Beetle”

• Center line 2 (product name): bold text “calming moisturizer”

• Bottom: italic motto “Your skin just fell in love”

Lighting: Side-lit from 45° with a tall softbox, plus subtle rim light behind both subjects to highlight micro-shadows, natural pores, and peach fuzz.

Camera: 80-MP medium-format, 100 mm macro lens, f/8, ISO 100, 1/125 s.

Background: Clean, neutral grey (±5% value), no props.

Output: Horizontal 1536x1024, 8K, uncompressed RAW look, editorial, unretouched.

Detail: Crisp skin texture, subtle freckling, full wardrobe visible for both models, well-lit eyes.

Color profile: sRGB, white balance 5400 K.

Film grain: 5%, no motion blur, no distortion.

NO airbrushed or waxy skin, no smoothing filter, no cartoonish/CGI look, no logo glitches, no duplicated figures or limbs, no cropped heads or clothing, no unrealistic plastic look, no spelling errors, no mirrored poses or compositions.

4 Likes

Beauty And The Beast

2 Likes

Almost got the perfect signage pop out of 4o…
It’s handling words way better than the start of the year :slight_smile:

7 Likes

I’m back baby!!! So sorry for my absence! I was too busy EscarGOAT’n… We can be a bit slow sometimes :snail::snail::snail:


:goat::spiral_shell:

:rofl::rofl::rofl:… no, I actually had a fever…

… and the only prescription was…

… more Bob Ross?!?… not what I expected, but it apparently worked :man_shrugging:t2:… feeling better than ever now, and eager to share some of the image projects I’ve been working on. WAAAY too much to share it all here, so hit me up if any of them catch your eye. I kinda go in “project” spurts, so plenty of a similar nature for the different types..

Super cool, and what an amazing idea! Way to translate this stuff into the real world, and I’m excited to hear how this goes for you! Keep us posted… speaking of post’n… whatcha’ll think of these?…

Pretty, pretty pretty, pretty Foresty, eh? EH???… :rofl::rofl::rofl:… oh well! My style is not for everyone…

… and that’s good! Iron sharpens iron!!!… anywho, it’s great being back, and I can’t wait to catch up on all of your amazing creations in this thread.

:victory_hand:

5 Likes

Well, it’s going kinda like this:


Chilling in a coffee house and pitching the ability to transform people into anything or anywhere…

7 Likes

Face Off

Geez, one can make movie characters with this tool…

3 Likes




4 Likes

…with 4o…

6 Likes

@polepole Woo! Those mockups would be very valuable for reproducing the same character. Good job. What are you prompting for the multi angle? Just a “mockup”?

3 Likes

There is probable a AI witch can generate a 3D model of this.

… with gpt-image-1 using input_fidelity=high…

Prompt

8 Likes

7 Likes

I don’t use just one prompt, it varies depending on the object, character in the image, or the image size. The prompt can be adjusted as needed. Also, the same prompt doesn’t produce identical results when used with the API vs ChatGPT.

You can provide a character portrait image and test it using the new input_fidelity=high parameter, although I haven’t tested that myself yet.

Using prompts with API




4 Likes

@polepole Sweeeeeet, thank you for sharing. “Four Identical Quadruplets” is very, very clever.

I’ve just been trying some of the new o3 image use via ChatGPT. We’re seeing MUCH BETTER consistency between illustrations with far less prompting or retries, check it out.

So I’m using dndGPT to build a Final Fantasy 6 tribute Dungeons and Dragons campaign with the new models. I’m seen much better consistency from one illustration to the next using o3 as the model. It surprises me by spending up to several minutes thinking of an image:

  • Also, check out the caption! This file was edited and re-exported from Photoshop when the branding was added, and now this AI tag originally generated by ChatGPT is accompanying it. Neat.

  • I also put this into Sora to try for some drama.

So to tell my story, I need Magitek Armor. So we built the stats and then did a quick mockup:

• It really nailed a “stained, sepia, yet still colored pencil look.”

Then, multiple requests following this illustration yielded pleasingly close results to the original (which I praised the model for creating).



Then went and did other things for awhile, and when I came back, even though I was in the same thread, the model seemed to have “lost it’s train of thought.”

So I referenced the sepia photo to retrain, and it seemed to work with far fewer shots than a similar process I was using last year. Though, last year, getting a “similar” illustration took several more tries. O3 had it immediately after retraining:


I continue to observe that repeating the style and dimensions in every prompt—i.e. “a nearly photorealistic rich colored pencil style in square dimensions”—keeps the results more consistent.

6 Likes

I put my images of Terra up on Facebook and started an Ai war in the comments. So many folk so jealous.

Nice pics, guy.

2 Likes

Make Scarecrows Great Again

1 Like

Whoops… I ran out of ink…

Anyone have a leftover Bob Ross prescription they could share?… I’m Jones’n for for color!

4 Likes

A cozy and humorous bathroom scene showing a dog pampered like a spa guest.


Intimate and tender portrait of a small dog freshly dried after a bath.

A playful, human-like depiction of a dog engaged in a common bathroom pastime.

Curious puppy waiting outside a steaming shower, eager yet patient.

7 Likes

Can I help out with some ink? :face_with_hand_over_mouth::cherry_blossom:

7 Likes

How could I say neigh to that :winking_face_with_tongue:

… Hey, I think it’s working! Thank you!!!… but I need MOHER!!!

… I’m just insatiable I guess :man_shrugging:t2:

6 Likes