Hang On

[Quicksand-Lover-7]

Okay I see, thanks.

[Viridian]

Is posting 3-5 variations of a concept fine, or is it just 1 image per concept.

To add to Dave's response, what we consider to be a "variation" isn't just every output you make from the same prompt. Our "variations" of a concept are created from separate attempts that illustrate significantly different outputs.

Using the current AI Challenge Thread theme, here are three different variations of the same concept.

business1.jpg

business2.jpg

buisness3.jpg

The concept is clear and consistent, but each image output is significantly different in composition. Each one was made with different source materials and combinations. They can convey different things. These are "variations" of a concept. While some people might love entire threads full of this, there is a point where they all end up looking the same.

What you describe as "variations" is more like alternate outputs from the same source or prompt:

buisness3.jpg

business5.jpg

business4.jpg

There's nothing really different here. I'm not even trying to swap anything out. I ran the same image through the same prompt.

What you were doing with the Hula Girl concept was 20 pictures of a hula girl on a muddy beach with a volcano in the background. One was enough; the other 19 offered nothing different. Pick your best one, not all of them.

[Viridian]

Also consider the level of effort put into AI work. The problem with the spammy outputs is that they are zero effort. Literally anyone can type in your prompt into ChatGPT right now and create an infinite amount of hula girls on a muddy beach with a volcano.

hula3.png

hula2.png

hula1.png

What distinguishes the AI artist is nuance in style and originality. A lot of us have unique workflows that use different models and apps. We can play with original characters and settings, and no prompt on ChatGPT will get the style and texture that I use in my work, for example.

While advances in AI app platforms make it even easier to create high quality images, because they are streamlined and watered down for casual users, they all end up looking the same.

[Viridian]

Sounds good. Credit given where credit due has always been the way.

I have a lot to say about this, but I'd just be stirring the pot. Just know that not everyone is OK with their work being used for AI creations, even other AI creators, and the QSF rules don't align with this. It's more about respect than rules. Be mindful of this is if you're going to use others' work as source images - AI or human alike.

Also, I'd refine your understanding of fair use. Firstly, fair use is a defense - it only becomes relevant should someone sue you for copyright infringement, and independent artists aren't going to front up the time and money to do that. It doesn't mean you're legally right, and I'd argue that this use would not meet the factors of fair use. More importantly though, just because something isn't intended for commercial purposes doesn't make it immune to copyright law.

Providing proper credit for source material is where we're at as far as rules go.

While we're fixated on the credit/permission point, Rule #2 covers this:

AI renders must be in a different medium from copyrighted sources. For example photo-realistic renders cannot come from a copyrighted photo. Video cannot be rendered from a copyrighted video source. Renders in comic book style or Anime styles are ok, for example.

The issue here is how transformative a work is. The rule states that the output has to be of a different medium to the original (copyrighted) source.

The question I have is: does "medium" mean "style"? Not just turning an image into an animation, but also requiring effort be made to alter the visual style of the original source?

In my case, I create original realistic AI animations based on Acidtester's Safari Girl. I have permission to use his character for starters, but regarding the derivative work, AT isn't in the business of creating photos, photorealistic AI images or AI animation, so I'm in a completely different lane.

But is it allowable to just upload an original AT drawing and just let the AI animate it?

[dlodoski]

I'm not going to respond to heresay regarding other creators. If someone has a real beef with the existing rules, I'm not hard to find.

And, Fair Use is not just a defense. It guides many decisions before the fact.

But while the modern cases all say fair use is just a defense, it’s clear it’s more than that. Fair use is a critical part of how copyright lawyers and judges determine what is infringing in the first place, which is why copyright analyses often jump to it before infringement has been clearly established.

https://law.marquette.edu/facultyblog/2 ... e-or-both/

The rest of this comes down to the distiction between Medium and Style. The existing rule does not address this properly, and will be refined in due course.

...But is it allowable to just upload an original AT drawing and just let the AI animate it?

This is a misleading question, because the word 'just' conveniently leaves out the issue of giving credit - which is a requirement.

Other than that, the answer is yes.

[Viridian]

This is a misleading question, because the word 'just' conveniently leaves out the issue of giving credit - which is a requirement.

The requirement of giving credit is already a given assumption and not up for debate. It takes zero effort to give credit. That isn't the issue. I was addressing Rule #2, not Rule #1.

The issue I raised there is whether an AI submission needs to pass a transformative threshold. My question was whether you were expected to put effort into altering the form and/or style of the work, or whether you could "just" use the raw source image (with credit). I used the word "just" in a derogatory manner to describe zero-effort image-to-video outputs that might arise. More specifically, my example was talking about the creative workflow (i.e. you "just" upload the image file to the AI app and then click "Create"), not the part where you share it on QSF (which requires attribution to the original creator).

This is a discussion better held in the relative thread, so I will move my longer response there.

[Viridian]

I am seeking clarity on Rule #2:

AI renders must be in a different medium from copyrighted sources. For example photo-realistic renders cannot come from a copyrighted photo. Video cannot be rendered from a copyrighted video source. Renders in comic book style or Anime styles are ok, for example

I think my understanding of the intention of this rule has been lost over time. My current understanding sees this as a rule about transformative and derivative works, but this may have been written with the intent of preserving copyrighted work. I'm not sure anymore

Specifically, I'm seeking clarity on whether turning an image to a video was considered a different "medium", given that the examples seem to reference style rather than medium.

The rule states:

For example photo-realistic renders cannot come from a copyrighted photo.

A photo-realistic render of... what? An image render and/or a video render? If a photo is used as a source image, can you make a photo-realistic image and/or video from it?

The following line states:

Video cannot be rendered from a copyrighted video source.

While this CAN be done (some apps allow you to swap the style of a video, for example), that's not how most people create their AI animations. The easiest and most common way is either create a video from a text-to-video prompt, or use image-to-video. But can I use a still from a video for an image-to-video output? Can I use a still from a video to create a still image? Can I create a video from a video source in a different style (e.g. cartoon animation to realistic)?

This part confused me the most:

Renders in comic book style or Anime styles are ok, for example.

Renders of what in this style and in what medium? Is this saying you ARE allowed to produce an anime image from another anime image? Or is it saying that you can make an anime video from a photorealistic video?

I don't need these questions answered - I personally don't use copyrighted source material to create my AI work. However, it might help to rewrite the rule for added clarity, as we're seeing more diverse AI creations and the tools are becoming more accessible and easier to use.

[dlodoski]

Like I said, it's something that needs to be worked out.

viewtopic.php?f=30&t=29870&start=14

...I don't need these questions answered.

Noted.

[BogDog]

Renders in comic book style or Anime styles are ok, for example.


Renders of what in this style and in what medium? Is this saying you ARE allowed to produce an anime image from another anime image? Or is it saying that you can make an anime video from a photorealistic video?

I think the idea behind barring photo-realistic video made from existing photo-realistic video but not 'toons or anime is the prevention of fraud. The issue of using existing video of people, especially celebrities and politicians, to make new photo-realistic video that delivers a different or even completely opposite message than the original did is the biggest concern. Perhaps video of a speech altered to deliver a different message.

An anime video rendered from a real one will not be a concern for such criminal fraud, which is crime prevention's biggest concern.

[Viridian]

That doesn't make sense to me. If the forum is making a stance against use of real figures and deep fakes, the rules should specify that the use of real people or photorealistic images are not permitted.

The rule specifically outlines the difference in medium, implying that video to video is not allowed, but image to video is. You can make a deep fake using an image source.

Nor do I really see why people would be making deep fakes on political speeches... on a quicksand fetish forum.