Toksik: forum

null

Additional threads to help you all start the game:

Baldrs new players advice guide for new players

Happy jump guide for new players for gaining stats efficiently! 

No, I spelled custom AI-generated graphics correctly.

“AI slop” is a criticism of low-effort, indiscriminately generated material: hundreds of barely inspected images produced from generic prompts, containing obvious anatomical errors, incoherent objects, broken text and no meaningful creative direction. It does not automatically describe every image produced using generative technology.

Calling every AI-assisted image “slop” is rather like calling every Photoshop creation “a filter,” every photograph “a camera preset,” or every digital illustration “tracing.” It confuses the tool with the quality of the finished work.

I am not advertising the ability to type “cool gangster woman, pink background, highly detailed, 4K” into a generator and sell whatever appears first. I am offering a controlled service involving concept development, prompt construction, model selection, reference-image preparation, generation settings, composition testing, variation assessment, regional corrections, inpainting, outpainting, upscaling, colour correction, sharpening and final presentation.

A customer gives me an idea. I then have to translate that idea into visual instructions that a model can interpret, identify why the initial outputs fail, correct those failures and continue refining the image until it actually resembles what was requested.

That process can include:

🧠 Converting a vague customer request into a workable visual concept.

🎨 Establishing a coherent colour palette, lighting direction, mood and composition.

🖼️ Choosing or creating appropriate reference images.

⚙️ Selecting the suitable model, checkpoint, LoRA, sampler, scheduler, resolution, denoising strength, guidance settings and generation workflow.

📐 Controlling pose, depth, edges, perspective or composition through conditioning tools.

🔍 Examining multiple generations for malformed hands, asymmetrical eyes, broken clothing, fused objects, inconsistent jewellery, impossible backgrounds and other artefacts.

🛠️ Masking and regenerating individual regions instead of discarding the entire image.

✂️ Removing unwanted objects, extending the canvas and rebuilding damaged areas.

🌈 Applying colour correction and selective grading so that separate generated elements appear to belong in the same image.

✨ Refining contrast, detail, texture, highlights, shadows and focal emphasis.

📏 Preparing the result for the customer’s required dimensions rather than merely stretching or crushing an image into the space.

🔁 Repeating this process until the image is suitable for delivery.

The existence of bad AI imagery does not establish that good AI imagery requires no skill. It establishes precisely the opposite: the enormous difference between uncontrolled and controlled results demonstrates the value of knowing what you are doing.

This is not simply my personal attempt to make the process sound complicated, either. ControlNet—the technology commonly used to constrain pose, depth, edges, segmentation and structure—was introduced specifically because ordinary text prompting does not provide sufficient spatial control. Its research describes conditioning diffusion models using structural information such as human pose, depth maps and edge detection. In other words, researchers literally developed additional neural-network architecture because typing a sentence and hoping for the best was not enough.

Professional production systems now represent generative imaging as a workflow, not a magic prompt box. Adobe’s own production tooling uses connected processing nodes, batch operations, source materials and iterative refinement. Its documentation specifically describes building custom workflows by combining and adjusting different nodes to create and revise assets.

Adobe’s Generative Fill also combines AI generation with manual selection, brush coverage, edge control and regional editing. The person operating it decides exactly which area should change, how large the affected region should be and how softly it should blend into the surrounding image. That is directed editing, not autonomous one-click image creation.

Major professional organisations and agencies—including Accenture, Dentsu, IPG Health, Publicis, Monks, PepsiCo/Gatorade and Estée Lauder—have used generative tools within actual creative-production workflows. Regardless of whether somebody personally likes AI imagery, pretending the entire field consists exclusively of teenagers pressing a “generate” button is demonstrably outdated.

Adobe has also documented uses by recognised brands and professional production groups for rebranding, campaign development and asset creation. These businesses are not replacing their entire creative departments with somebody entering “make advert good” into a website. They are integrating generation into systems that still require art direction, brand consistency, review, correction and final approval.

None of that means every generated image is good.

Most are not.

Most photographs are not professional photography. Most drawings are not professional illustration. Most Photoshop edits are not professional compositing. Most pieces of writing are not professional copywriting. The accessibility of a tool increases the quantity of mediocre output, but it does not eliminate the expertise involved in producing the better output.

Photoshop itself makes an excellent comparison.

Opening Photoshop is easy. Applying a preset is easy. Dropping text over an image is easy. Using an automated action is easy. None of those actions makes somebody a skilled designer. Skill becomes visible through selection, composition, masking, typography, lighting, colour, visual hierarchy and the ability to diagnose why something looks wrong.

The same principle applies here.

Opening an image generator is easy.

Producing an image that consistently follows a specific brief is not.

The generator does not automatically understand that the customer wants a particular facial structure maintained across revisions. It does not reliably understand which objects must remain untouched. It does not know that a background detail conflicts with the intended era. It does not know whether a particular composition will work inside a 600×100 Torn signature. It does not know that the focal character will be hidden once typography and borders are added. It does not know which generation defects a paying customer will immediately notice.

The operator has to recognise and correct those problems.

There is also a fundamental difference between generation and direction.

A camera generates a photograph by capturing light. That does not make the photographer irrelevant.

A synthesiser generates sound electronically. That does not make the musician irrelevant.

A 3D renderer generates the final pixels of a scene. That does not make the modeller, lighting artist or compositor irrelevant.

Photoshop calculates blending modes, selections, transformations, filters and content-aware corrections. That does not mean Photoshop created the design independently.

Likewise, an AI model performs the mathematical synthesis of an image, while the operator directs, selects, constrains, rejects, edits and presents the result.

The machine provides possibility.

The human provides intention and judgement.

Those two things are not interchangeable.

The term “AI slop” is useful when describing mass-produced content that nobody cared enough to inspect. It becomes meaningless when lazily applied to every possible use of generative imagery. At that point, it is no longer criticism of quality. It is merely a reflexive label being used in place of actually evaluating the work.

Judge an image on whether it looks good.

Judge it on whether it matches the brief.

Judge it on its composition, consistency, atmosphere, detail and presentation.

Judge whether the customer believes it was worth the agreed price.

But saying “AI slop” before even seeing the finished product is not artistic criticism. It is just announcing that you have already decided which tools people are permitted to use.

I am charging $5,000,000 in Torn currency, not claiming to be producing a hand-painted Renaissance ceiling. The customer provides the concept, I develop and refine the visual, and they receive a custom image at a very modest in-game price.

Nobody is being deceived about the production method. “AI-generated” is written directly in the advertisement.

Nobody is being forced to purchase it.

Nobody is being prevented from hiring a traditional graphic artist.

Nobody is claiming that AI and hand illustration are identical disciplines.

I am simply offering a clearly described service to people who may want it.

The funniest part is that if I generated an image badly, accepted the first result, ignored obvious defects and uploaded fifty nearly identical variations, then “AI slop” might be a fair description.

But the fact that bad results are possible is exactly why model control, careful selection, inpainting, refinement and editing matter.

The insult therefore does not disprove my advertisement.

It accidentally proves its central point.

There are people who press Generate once and accept whatever comes out.

There are also people who understand the tools, construct controlled workflows, reject unsuitable results and refine the surviving image into something coherent.

I am advertising the second service.

You are free not to buy it, naturally.

However, repeatedly calling an entire production method “slop” does not make you an art critic. It merely saves you from having to offer any specific criticism of the actual image.

So no—I did not spell “AI slop” incorrectly.

You simply mistook the name of a medium for a judgement of quality. 😏

fixed

Nice work, good size. 

Almost at 100 R- 

Lets get it.

b

Here for the eventual 100 R- 

R- 

R-

How many threads does a cockwomble need? 

That's a lot of words to say you're a spastic. 

must have been real hard to type words into the GenAI bro 

Not bad, good size. 

Thread joke tickled me. 

wow, u must have amazing mentor with such improve 

i dont see any problem here 

i can work with ai in all capacities. a lot of manual editing takes place when my paid generations are used to make sure they are highest quality. such as inpainting, color corrections and more and that needs to be manaual i would say its more effort than just finding a image on google or other web pages and slapping it on with no thought i can make things entirely custom. i said it is manual editing in my original post. people thought to troll despite selling ai generated base work themselves in the communities i come from if i did that i would be banned without disclaimers on everyhing if you use one ai image your whole work is ai and the peoiple should know 

then why are you all using ai images in your art too and not telling your buyers almost every signature in this forum is ai as a base. its only bad if it is declared and someone can make there own generated image before making it into a signature with manual editing??

the tutorial thread says you have to provide information if you use ai and i use ai  this person wanted a render so i used a render if another person wants a ai generated image first then i will create it and use that 

this is manually made in photoshop with a provided render from gmap and the file was provided to him for his review

sorry to disappoint you

good artistry rated positively 

low quality input, low quality output