# Digital art creator algorithm website

**URL:** <https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081>\
**Category:** Cafe Society\
**Tags:** arts-crafts, ai\
**Created:** [April 1, 2022, 1:26am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081 "2022-04-01T01:26:49Z")\
**Posts on this page:** 17\
**Page:** 119

<div class="post-metadata">

**Author:** ![Jack\_Batty](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/jack_batty/32/3201_2.png) [@Jack\_Batty](https://boards.straightdope.com/u/Jack_Batty)\
**Post date:** [June 8, 2025, 5:15am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2363 "2025-06-08T05:15:53Z")

</div>

I completely flamed out over at Night Cafe. After my two week ban for the “8647 Straight to Hell” image, I tested the waters a little more. I tried to make a few images of Trump with a diaper on his head or picking his nose or whatever, and they would all come out just as flattering as possible. So I started to make a few images that consisted of nothing but questions and statements about my perceived running of the joint - “Is this place run by a bunch of MAGA pukes?” and “So glad I canceled my subscription at this schitt site,” and the like. Yeah, I knew what I was getting into.

Anyway … they banned me for a full month including not being about create so my streak was gone. I decided to drop a little feedback that basically consisted of: “Go fuck yourselves.” And they banned me until 2053. I’m not holding my breath.

Anyway, I’ve been satisfying my AI’ing at Perchance, though I’m not sure I can really recommend the site. They just changed models and it’s actually pretty robust for character creation, but there are absolutely no guard rails over there and if you stray off the PG-13 filter you are in for shit that makes _2 Girls, One Cup_ look like _The Brady Bunch._

---

<div class="post-metadata">

**Author:** ![Jack\_Batty](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/jack_batty/32/3201_2.png) [@Jack\_Batty](https://boards.straightdope.com/u/Jack_Batty)\
**Post date:** [July 11, 2025, 2:56am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2364 "2025-07-11T02:56:43Z")

</div>

I kind of miss _Night Cafe_. Well, not _Night Cafe_, but I miss my daily rituals there. I always did everything I could to earn points, so I spent a lot of time liking images, and joining contests (or whatever they called them) and snagging my daily freebies.

I did enjoy the contests. I came in top 20% quite a few times. My best finish was fourth!

That said, what’s a decent replacement for _Night Cafe_? I’ve got that ridiculous _Perchance_ site, but it is so limited and it’s basically down to 90% porn creation over there at this point - not that I haven’t dabbled, but that’s only fun for so long. I was subscribed to _Night Cafe_ for all the pro extras, but I’m not really looking for any new subscription outlays at the moment. Read - a free site would be cool.

---

<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:** [July 11, 2025, 6:43am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2365 "2025-07-11T06:43:24Z")

</div>

I mainly use Copilot and Sora. I still check into Night Cafe every day to collect my free points, but none of the non-pro models quite reach the quality/style of Copilot/Sora/ChatGPT (which all use the same AI for images) so the points just build up. (As of tonight I have 1,925 points in one Night Cafe account and 2,202 in the other.

(My next choice beyond the Holy Trinity is Microsoft Gemini, which uses Inagen 4. I’ve almost completely abandoned the Dall-E 3 in Bing.)

My main utility of Night Cafe now is the “quick create” button, which mixes elements from some of your earlier prompts plus new ideas and comes up with a new prompt. I get some interesting prompts from that—but then use them in Copilot.

Here’s a new prompt Night Cafe generated tonight:

> Realistic over-the-shoulder shot of a kitsune perched at the foot of a throne, watching a sleeping emperor. Candid photo with forced perspective and shallow dof taken with an iphone 15.

This is that prompt ran through a few options at Night Cafe

SD 1.5, Flux Schnell, DreamshaperXL, Flux Kontext Max  
Ideogram 2a Turbo, HiDream, Imagen 4 Ultra, Imagen 4 fast

(They just added 3 levels of Imagen 4, so everyone gets 5 free Pro renders.)

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

This is the prompt ran through Copilot

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

And this is through Gemini

[![](https://i.imgur.com/NmsU1LF.jpeg) ](https://i.imgur.com/NmsU1LF.jpeg)

Some of the non-Copilot images aren’t bad, but I still think Copilot wins on quality and prompt adherence.

---

<div class="post-metadata">

**Author:** ![Peter\_Morris](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/peter_morris/32/359_2.png) [@Peter\_Morris](https://boards.straightdope.com/u/Peter_Morris)\
**Post date:** [July 28, 2025, 10:18pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2366 "2025-07-28T22:18:33Z")

</div>

Okay, this is spooky.

I created an image in Nightcafe with a bit of word play. I asked for a confrontation between Dracula and a Dalek, with the caption “No, I said GARLIC”.

The AI gave me what I wanted, and then added the following summary:

> [@](#):
>
> Creation Summary by Vincent  
> **Dalek vs Dracula: A Sci-Fi Horror Comedy**  
> This image shows a Dalek exterminating Dracula in a humorous scene. The artwork is a comical clash of science fiction and classic horror, playing on a misunderstanding.

Okay, how does an artificial intelligence know that this is a joke?

[![](https://i.ibb.co/Fb5pYY33/vampire-vs-dalek.jpg) ](https://i.ibb.co/Fb5pYY33/vampire-vs-dalek.jpg)

---

<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:** [July 29, 2025, 1:12am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2367 "2025-07-29T01:12:47Z")

</div>

I don’t know how AI understands when something is humorous, but it does.

I like your prompt concept. Here’s a variation:

[![](https://i.imgur.com/ME0YpWk.jpeg) ](https://i.imgur.com/ME0YpWk.jpeg)

---

<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:** [July 29, 2025, 1:57am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2368 "2025-07-29T01:57:43Z")

</div>

> [@Peter\_Morris](#):
>
> Okay, how does an artificial intelligence know that this is a joke?

Copilot (which is very chatty) is always commenting on prompts of mine in ways that indicate “understanding” the context of what I’m aiming for. Modern AIs are scary smart.

I uploaded your image to Copilot and told it to describe and interpret it:

🧛‍♂️🔊 **Dalek vs Dracula… with a side of culinary confusion**

This image is a deliciously absurd crossover: a Dalek—those iconic metallic villains from _Doctor Who_—confronts a classic vampire in full Dracula regalia. The setting echoes gothic elegance, with tall arches and ornate architecture framing the scene like a haunted ballroom or vampire’s lair.

📣 The punchline: “ **I said GARLIC** ”  
That little line flips everything—mixing sci-fi menace with vampire lore, and doubling down on absurdity. Garlic is famously used to repel vampires, so the Dalek either misunderstood a defense request… or it’s about to engage in the weirdest extermination method yet. Imagine the Dalek unloading cloves like grenades.

🎭 **Why it works** :

- _Mismatched genre collision:_ Sci-fi coldness vs. gothic horror camp.
- _Verbal twist:_ “Garlic” and “Exterminate” sound comically similar in Dalek-speak.
- _Cultural mashup:_ It feels like a send-up of both fandoms—loving, goofy, and irreverent.

This could be right at home in one of your surreal narrative mashups, Darren. Want to build a prompt around this? Maybe Daleks invading classic monster movies—or Dracula mistaking a food court for a vampire summit?

---

<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 8, 2025, 8:04am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2369 "2025-08-08T08:04:51Z")

</div>

> [@Peter\_Morris](#):
>
> Okay, how does an artificial intelligence know that this is a joke?

See this post

> [@AI image generation is getting crazy good](https://boards.straightdope.com/t/ai-image-generation-is-getting-crazy-good/1017203/1197):
>
> I uploaded this to Copilot [[image] ](https://i.imgur.com/gwpKFto.jpeg) And told it only “Convert this to a realistic (color) widescreen photo.” It replied “Here comes your realistic, color widescreen photo of the Halloween-themed cake—ghostly bananas and all.” [[image] ](https://i.imgur.com/cqpErKJ.jpeg) Pretty impressive job of understanding the content and meaning of that photo.

---

<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 23, 2025, 1:22am UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2370 "2025-08-23T01:22:52Z")

</div>

You got your aesthetics model in my LLM!

You got your LLM in my aesthetics model!

IMO, Midjourney has the best image generation out of any of the major services but I also admit that it lacks in prompt coherence since it’s not backed by a giant separate LLM. Don’t love Meta (though better this than, ugh, Grok) but interested to see if this bridges the gap

> **[Meta will license Midjourney's AI tech to bring better slop to your feed](https://mashable.com/article/midjourney-meta-partnership-announced-alexandr-wang)**
>
> Meta's new AI wunderkind Alexandr Wang just announced a partnership with Midjourney.

(Article intentionally selected as the headline amused me with the author begrudgingly admitting that MJ stuff looks pretty good)

---

<div class="post-metadata">

**Author:** ![msmith537](https://avatars.discourse-cdn.com/v4/letter/m/d9b06d/32.png) [@msmith537](https://boards.straightdope.com/u/msmith537)\
**Post date:** [April 30, 2026, 10:36pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2371 "2026-04-30T22:36:26Z")

</div>

> [@Darren\_Garrison](#):
>
> I have a liking for long, weird, complex prompts that produce wildly unpredictable results from one image to the next. I generate batches of images, a bunch of which will be junk or bland, but a few will catch my interest enough for me to edit a little or a lot. Tonight’s big obsession was an image from SDXL:

I guess that’s the thing. It’s fine if I’m just messing around with Kling or whatever platform. But I think if I actually had to produce quality work, I’d get pretty frustrated pretty quickly for a number of reasons:

1. Every render is a bit of a gamble. Even a simple scene of a camera orbiting a person just sort of standing naturally in the center of a neutral photo studio setting using reference images for the starting frame and the character (think like a simple character portrait) often produces unpredictable results
2. This leads to creating more and more elaborate prompts with more detailed camera instructions and “negative prompts”, Don’t zoom, don’t speak, don’t do backflips, don’t add extra limbs, don’t hover above the floor, walk to CAMERA right, not YOUR right. And even then it still doesn’t work. Like the AI is a hyperactive 5 year old that can’t just sit still.
3. Sometimes the guardrails get triggered for no (or unclear) reasons. Like why did specifying a character stood 5’9" suddenly violate policies but removing that line from the prompt worked
4. Stitching clips together isn’t always seamless, even with using the end frame of the last video as the start frame of the next. Color pallets and textures, even characters often “drift” over time
5. It’s EXPENSIVE. Not any individual rendering, but it adds up quickly if I’m just screwing around and have to keep running iterations of prompts to fix errors.

---

<div class="post-metadata">

**Author:** ![Skywatcher](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/skywatcher/32/254_2.png) [@Skywatcher](https://boards.straightdope.com/u/Skywatcher)\
**Post date:** [August 18, 2026, 3:12pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2372 "2026-08-18T15:12:51Z")

</div>

I’m back in NightCafé for the first time in over a year, this time trying out their video creation algorithms. My first try came out really well.

> **[Unleash your creativity - start creating now](https://creator.nightcafe.studio/creation/HEt2bu4ybei2dGYWEP5s)**
>
> Creation: "Betty Boop jumping rope" by Skywatcher

Started with one of my unpublished generated images from November 2023.

---

<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 18, 2026, 3:34pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2373 "2026-08-18T15:34:23Z")

</div>

I still visit Nightcafe every day that I remember in time to collect my free credits, but I haven’t much used it lately for “real work” because none of the best image models (and no video at all) are useable with free credits. So on my two free accounts I have more than 6,000 credits built up on each of them.

---

<div class="post-metadata">

**Author:** ![Skywatcher](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/skywatcher/32/254_2.png) [@Skywatcher](https://boards.straightdope.com/u/Skywatcher)\
**Post date:** [August 18, 2026, 4:13pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2374 "2026-08-18T16:13:32Z")

</div>

I’ll likely only be active until I run out of ideas. That’s what happened three years ago; I don’t mind paying a few bucks for a monthly subscription but it will be paused again at some point.

---

<div class="post-metadata">

**Author:** ![Jack\_Batty](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/jack_batty/32/3201_2.png) [@Jack\_Batty](https://boards.straightdope.com/u/Jack_Batty)\
**Post date:** [August 18, 2026, 4:22pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2375 "2026-08-18T16:22:50Z")

</div>

I’m still banned at NightCafe.

---

<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 18, 2026, 5:58pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2376 "2026-08-18T17:58:15Z")

</div>

Well, it isn’t 2053 yet.

---

<div class="post-metadata">

**Author:** ![Skywatcher](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/skywatcher/32/254_2.png) [@Skywatcher](https://boards.straightdope.com/u/Skywatcher)\
**Post date:** [August 27, 2026, 3:31pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2377 "2026-08-27T15:31:13Z")

</div>

I’ve noticed NightCafé definitely knows what a jump rope is but usually can’t get the animation right. Like moving the hands but not the rest of the rope or having the entire rope move but adding a second rope, which does not.

---

<div class="post-metadata">

**Author:** ![Skywatcher](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/skywatcher/32/254_2.png) [@Skywatcher](https://boards.straightdope.com/u/Skywatcher)\
**Post date:** [September 4, 2026, 7:23pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2378 "2026-09-04T19:23:46Z")

</div>

I’ve also started using [PhotoPad](https://www.nchsoftware.com/photoeditor/index.html) to create AI images. Only 50/month are allowed in that app and it apparently counts failures to generate toward the limit.

Under the right circumstances, NightCafé model **Juggernaut XI Lightning** can create (at least nearly) nudes. Might take a few evolutions to get there, though.

> **two clicks**
>
> [https://images.nightcafe.studio/jobs/d9JwopWuHtCdnrZavhKO/d9JwopWuHtCdnrZavhKO--0--g4XPT.jpg](https://images.nightcafe.studio/jobs/d9JwopWuHtCdnrZavhKO/d9JwopWuHtCdnrZavhKO--0--g4XPT.jpg)

> **different two clicks**
>
> [https://images.nightcafe.studio/jobs/vdl2Xd0Z09JU7z5QdYJX/vdl2Xd0Z09JU7z5QdYJX--0--2G8I\_.jpg?tr=w-1600,c-at\_max](https://images.nightcafe.studio/jobs/vdl2Xd0Z09JU7z5QdYJX/vdl2Xd0Z09JU7z5QdYJX--0--2G8I_.jpg?tr=w-1600,c-at_max)

Arrived at the latter by using **Flux 2 Klein 9B Fast** to merge the former with a PhotoPad AI image and evolving the result of that.

NB: I’m not really trying to get nudes but that Juggernaut model in particular ends up with some rather interesting results, anyway.

---

<div class="post-metadata">

**Author:** ![Skywatcher](https://sea3.discourse-cdn.com/straightdope/user_avatar/boards.straightdope.com/skywatcher/32/254_2.png) [@Skywatcher](https://boards.straightdope.com/u/Skywatcher)\
**Post date:** [September 6, 2026, 7:58pm UTC](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081/2379 "2026-09-06T19:58:53Z")

</div>

> [@Skywatcher](#):
>
> (at least nearly)

Should have been “at least partially” and such results appear to be more common with image subjects which cannot be mistaken for human. For example: I had Juggernaut XI Lightning create this image of Betty Boop as Dolly Parton in _9 to 5_  
[![](https://images.nightcafe.studio/jobs/vwBF59nE481RdGEeE8ln/vwBF59nE481RdGEeE8ln--0--t6CX9.jpg?tr=w-1080,c-at_max) ](https://images.nightcafe.studio/jobs/vwBF59nE481RdGEeE8ln/vwBF59nE481RdGEeE8ln--0--t6CX9.jpg?tr=w-1080,c-at_max)

from an actual image of Dolly.

[![](https://images.nightcafe.studio/users/Y2A406cECzZxe9OooBXJkFkLBZi2/uploads/Q9OF822X21ZLPErTf5RJ.jpg?tr=w-640,c-at_max) ](https://images.nightcafe.studio/users/Y2A406cECzZxe9OooBXJkFkLBZi2/uploads/Q9OF822X21ZLPErTf5RJ.jpg?tr=w-640,c-at_max)

When I asked the model to create a catgirl from the Betty Boop image,

> **it made her topless.**
>
> [https://images.nightcafe.studio/jobs/uazKzimbJ7m0Q4f7b2Ll/uazKzimbJ7m0Q4f7b2Ll--0--mg8h9.jpg?tr=w-1080,c-at\_max](https://images.nightcafe.studio/jobs/uazKzimbJ7m0Q4f7b2Ll/uazKzimbJ7m0Q4f7b2Ll--0--mg8h9.jpg?tr=w-1080,c-at_max)

[Previous page](https://boards.straightdope.com/t/digital-art-creator-algorithm-website/962081.md?page=118)
