# AI image generation is getting crazy good

**URL:** <https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203>\
**Category:** Miscellaneous and Personal Stuff I Must Share\
**Tags:** ai\
**Created:** [April 19, 2025, 3:22am UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203 "2025-04-19T03:22:52Z")\
**Posts on this page:** 20\
**Page:** 62

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [August 12, 2025, 10:07am UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1221 "2025-08-12T10:07:26Z")

</div>

Another prompt test:

Create a realistic photo of a Geronticus eremita attempting to lift a Dactylotum bicolor out of an empty wineglass on a 1970s linoleum kitchen floor.

As usual, Copilot is top of the heap

[![](https://i.imgur.com/MAxtcej.jpeg) ](https://i.imgur.com/MAxtcej.jpeg)

Next, the clueless: SDXL, Flux Schnell, Ideogram, and HiDream

[![](https://i.imgur.com/i6GDyiP.jpeg) ](https://i.imgur.com/i6GDyiP.jpeg)

And the not bad Imagen

[![](https://i.imgur.com/64d2Ze8.jpeg) ](https://i.imgur.com/64d2Ze8.jpeg)

---

<div class="post-metadata">

**Author:** ![Chronos](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/chronos/32/134_2.png) [@Chronos](https://boards.straightdope.com/u/Chronos)\
**Post date:** [August 12, 2025, 12:45pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1222 "2025-08-12T12:45:43Z")

</div>

One more thought on the waterfires: The best examples all appear to have disconnected bits. This makes it look more like both water and fire, since water will have loose droplets, and fire will have embers. Plastic or blown glass, though, will all be one connected piece, so the ones with one connected piece look more plasticky or glassy.

---

<div class="post-metadata">

**Author:** ![pulykamell](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/pulykamell/32/3166_2.png) [@pulykamell](https://boards.straightdope.com/u/pulykamell)\
**Post date:** [August 12, 2025, 1:53pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1223 "2025-08-12T13:53:17Z")

</div>

This is the best I could do with “fire made of water”: (ChatGPT 5)

[![](https://chatgpt.com/backend-api/public_content/enc/eyJpZCI6Im1fNjg5YjQ3MTExYjdjODE5MWIyNDU0NWQzZWQzMDY0MGM6ZmlsZV8wMDAwMDAwMGQxNGM2MWZiYWRjMDYxMDJkNDhiMTE3MyIsInRzIjoiNDg3NTAxIiwicCI6InB5aSIsInNpZyI6IjhmMDkyYjczODg3YzYwYWVjZDY1MmI1MTY5ZTQwNDIzNWMzOGUwY2Y3MWYxNWZlNjc5YjllYTY4ODM1YjM1ZGUiLCJ2IjoiMCIsImdpem1vX2lkIjpudWxsfQ==) ](https://chatgpt.com/backend-api/public_content/enc/eyJpZCI6Im1fNjg5YjQ3MTExYjdjODE5MWIyNDU0NWQzZWQzMDY0MGM6ZmlsZV8wMDAwMDAwMGQxNGM2MWZiYWRjMDYxMDJkNDhiMTE3MyIsInRzIjoiNDg3NTAxIiwicCI6InB5aSIsInNpZyI6IjhmMDkyYjczODg3YzYwYWVjZDY1MmI1MTY5ZTQwNDIzNWMzOGUwY2Y3MWYxNWZlNjc5YjllYTY4ODM1YjM1ZGUiLCJ2IjoiMCIsImdpem1vX2lkIjpudWxsfQ==)

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [August 13, 2025, 1:45am UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1224 "2025-08-13T01:45:01Z")

</div>

I had a sudden inspiration from the title of the Pet pictures thread…

> A middle-aged woman in a sweatsuit power-walking on a sidewalk in a suburban neighborhood. She has three framed photos on leashes (like small dogs on leashes running with her). Iphone 15 photo with shallow dof.

[![](https://i.imgur.com/qv6ExID.jpeg) ](https://i.imgur.com/qv6ExID.jpeg)

It understood everything pretty well except for the number three.

---

<div class="post-metadata">

**Author:** ![Tibby](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/tibby/32/17030_2.png) [@Tibby](https://boards.straightdope.com/u/Tibby)\
**Post date:** [August 13, 2025, 4:01pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1225 "2025-08-13T16:01:36Z")

</div>

Well, looks like I have to bail my cat out again.

[![](https://i.imgur.com/k2QWC64.jpeg) ](https://i.imgur.com/k2QWC64.jpeg)

---

<div class="post-metadata">

**Author:** ![Maserschmidt](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/maserschmidt/32/18829_2.png) [@Maserschmidt](https://boards.straightdope.com/u/Maserschmidt)\
**Post date:** [August 13, 2025, 4:31pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1226 "2025-08-13T16:31:37Z")

</div>

21 years old, the judge should be cutting that cat some breaks.

---

<div class="post-metadata">

**Author:** ![Tibby](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/tibby/32/17030_2.png) [@Tibby](https://boards.straightdope.com/u/Tibby)\
**Post date:** [August 13, 2025, 4:52pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1227 "2025-08-13T16:52:00Z")

</div>

ChatGPT got the DOB wrong. She’s only 2.

---

<div class="post-metadata">

**Author:** ![pulykamell](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/pulykamell/32/3166_2.png) [@pulykamell](https://boards.straightdope.com/u/pulykamell)\
**Post date:** [August 13, 2025, 5:02pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1228 "2025-08-13T17:02:08Z")

</div>

Also one gradation between 7” and 8”, but two between 8” and 9”. (And never mind the unrealistic measurements.)

---

<div class="post-metadata">

**Author:** ![Jophiel](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/jophiel/32/66_2.png) [@Jophiel](https://boards.straightdope.com/u/Jophiel)\
**Post date:** [August 13, 2025, 5:02pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1229 "2025-08-13T17:02:40Z")

</div>

“Where did this catnip come from?!”

_I got it from YOU, Dad!_

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [August 13, 2025, 5:40pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1230 "2025-08-13T17:40:14Z")

</div>

The most attractive Hobbes

[![](https://i.imgur.com/yFjnjaF.jpeg) ](https://i.imgur.com/yFjnjaF.jpeg)

> Hobbes the tiger from Calvin and Hobbes, except he’s very attractive. Iphone 15 photo with shallow dof and forced perspective. 9:16

---

<div class="post-metadata">

**Author:** ![Maserschmidt](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/maserschmidt/32/18829_2.png) [@Maserschmidt](https://boards.straightdope.com/u/Maserschmidt)\
**Post date:** [August 13, 2025, 5:41pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1231 "2025-08-13T17:41:37Z")

</div>

You monster.

---

<div class="post-metadata">

**Author:** ![Chronos](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/chronos/32/134_2.png) [@Chronos](https://boards.straightdope.com/u/Chronos)\
**Post date:** [August 13, 2025, 5:57pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1232 "2025-08-13T17:57:23Z")

</div>

Was that entirely AI-generated, in one pass? I’m a little surprised that it made the sign identical in both views. That’s the sort of thing that these generators often have difficulty with.

---

<div class="post-metadata">

**Author:** ![Tibby](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/tibby/32/17030_2.png) [@Tibby](https://boards.straightdope.com/u/Tibby)\
**Post date:** [August 13, 2025, 6:15pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1233 "2025-08-13T18:15:23Z")

</div>

No, it messed up the first sign, so I duplicated the second and pasted onto the first. Didn’t notice the wrong DOB, though.

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [August 13, 2025, 6:20pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1234 "2025-08-13T18:20:01Z")

</div>

Back on the relative scale issue, the ChatGPT/Copilot/Sora renderer seems to have a pretty good grasp that elephants are big.

[![](https://i.imgur.com/cyysK8F.jpeg) ](https://i.imgur.com/cyysK8F.jpeg)

[![](https://i.imgur.com/fLqF5cW.jpeg) ](https://i.imgur.com/fLqF5cW.jpeg)

[![](https://i.imgur.com/pu7XHJv.jpeg) ](https://i.imgur.com/pu7XHJv.jpeg)

Sample prompt

> **Summary**
>
> Realistic profile photo of an elephant standing in a crowded room. A wholesome 1950s businessman is writing a mailing address on the elephant. Iphone 15 photo. 16:9

(I also find it interesting what it comes up with for mailing addresses when left with making that decision.)

---

<div class="post-metadata">

**Author:** ![Darren\_Garrison](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/darren_garrison/32/92_2.png) [@Darren\_Garrison](https://boards.straightdope.com/u/Darren_Garrison)\
**Post date:** [August 13, 2025, 9:21pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1235 "2025-08-13T21:21:23Z")

</div>

Sora had no problem with the prompt “[MTG and AOC fighting with space lasers](https://boards.straightdope.com/t/most-attractive-hobbies/1022010/26)”.

[![](https://i.imgur.com/wKtjBn6.jpeg) ](https://i.imgur.com/wKtjBn6.jpeg)

But didn’t “get” MTG playing M:TG

[![](https://i.imgur.com/hH2eggi.jpeg) ](https://i.imgur.com/hH2eggi.jpeg)

---

<div class="post-metadata">

**Author:** ![Ponderoid](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/ponderoid/32/19221_2.png) [@Ponderoid](https://boards.straightdope.com/u/Ponderoid)\
**Post date:** [August 13, 2025, 9:37pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1236 "2025-08-13T21:37:01Z")

</div>

Yes, here it got the idea that the phone booth should be bigger than a 6-year-old kid, but… not nearly big enough.

Prompt: Draw a 6-year-old boy in a full sized phone booth.

[![](https://i.gyazo.com/2eec524e2bebe7efe275508a38462cbe.png) ](https://i.gyazo.com/2eec524e2bebe7efe275508a38462cbe.png)

Prompt: Draw a 4-year-old boy in a full sized phone booth

[![](https://i.gyazo.com/fcd4d77ba59aaaa1257ccd6f3da8b29b.png) ](https://i.gyazo.com/fcd4d77ba59aaaa1257ccd6f3da8b29b.png)

---

<div class="post-metadata">

**Author:** ![Chronos](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/chronos/32/134_2.png) [@Chronos](https://boards.straightdope.com/u/Chronos)\
**Post date:** [August 13, 2025, 9:58pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1237 "2025-08-13T21:58:33Z")

</div>

Wait, Marjorie Taylor Greene is actually Jace the Mind-Sculptor? That explains so much!

---

<div class="post-metadata">

**Author:** ![Maserschmidt](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/maserschmidt/32/18829_2.png) [@Maserschmidt](https://boards.straightdope.com/u/Maserschmidt)\
**Post date:** [August 13, 2025, 10:08pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1238 "2025-08-13T22:08:39Z")

</div>

Out of curiosity, I asked CoPilot to estimate how much energy it was using to generate one picture, quoted to me in the units of “time a standard microwave is on.” It danced around quite a bit about not being sure and data not being public, but eventually its answer was that depending on image complexity, generating one picture for me was like running the microwave for 30-50 seconds.

I then asked it how much energy it used in that calculation, and it said 1-7 microwave seconds.

No wonder they’re building big energy hubs!

---

<div class="post-metadata">

**Author:** ![GESancMan](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/gesancman/32/2923_2.png) [@GESancMan](https://boards.straightdope.com/u/GESancMan)\
**Post date:** [August 14, 2025, 3:44am UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1239 "2025-08-14T03:44:06Z")

</div>

> [@Darren\_Garrison](#):
>
> Sora had no problem with the prompt “[MTG and AOC fighting with space lasers](https://boards.straightdope.com/t/most-attractive-hobbies/1022010/26)”.

Got the colors of the lightsabers backwards, though.

---

<div class="post-metadata">

**Author:** ![Jophiel](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/jophiel/32/66_2.png) [@Jophiel](https://boards.straightdope.com/u/Jophiel)\
**Post date:** [August 14, 2025, 12:26pm UTC](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1240 "2025-08-14T12:26:27Z")

</div>

> [@Ponderoid](#):
>
> Yes, here it got the idea that the phone booth should be bigger than a 6-year-old kid, but… not nearly big enough.

Midjourney did pretty well with this, just prompting “Six year old in a phone booth”

[![](https://i.imgur.com/kZooAYg.jpeg) ](https://i.imgur.com/kZooAYg.jpeg)

Not exactly “nailed it” but two or three of my single test came out with pretty normal sized booths. I like the kid talking on a cell phone but, hey, he couldn’t reach the pay phone!

I used a couple local SDXL models and they made much smaller booths

[![](https://i.imgur.com/bOK4Ghz.png) ](https://i.imgur.com/bOK4Ghz.png)

[![](https://i.imgur.com/smpTE8O.png) ](https://i.imgur.com/smpTE8O.png)

I like the kid wearing headphones and trying to dial a door keypad.

Unrelated to children and phone booths, here’s my Mutants & Masterminds character, Molly Dynamite

[![](https://i.imgur.com/JQpIBF6.png) ](https://i.imgur.com/JQpIBF6.png)

[Previous page](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203.md?page=61)

[Next page](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203.md?page=63)
