Tag suggestion: image_response

I'm thinking of image_response as meaning exactly what it means on pixiv: an image created as a "response" to a prior image, one of the most notable examples being where someone draws an uncolored version of an image and various artists do their own rendition of it. It's also a familiar wording because even YouTube, I believe, has "video responses". That's what they're called, right?

Came across it most recently with post #509670. Couldn't figure out why it had two artist tags, and then realized why. Here's the original and three responses - link

Maybe it could also be used for pixiv projects like "expressions" and "hairstyles of my wife" and "let's make touhou kids", etc. I've only really thought it through in terms of the type of example I linked, though. So there could be downsides to other usages, and we might want to restrict it more, etc etc.

I'm not sure yet so that's why I'm asking. Thoughts?

That would just move danbooru closer to being a pixiv mirror, tbh. I don't see this as needing to be something expressed outside of comments.

It might be useful for explaining why certain images have more than one artist tags. My concern with this tag is that if the tag were searched for, it would contain miscellaneous images that really don't relate to one another other than the fact that they're all variations of an image.

0xCCBA696 said:
That would just move danbooru closer to being a pixiv mirror, tbh. I don't see this as needing to be something expressed outside of comments.

This is a somewhat unavoidable phenomenon.

I brought up a related issue in forum #24386 not too long ago, although it never reached any sort of conclusion.

I personally think the parent/child system does a good enough job already of crediting the original artist (by being the main parent) and showing which ones are responses (by being a child), but most people seemed against that.

I guess a tag could work, although it would just end up being a large collaboration of unrelated images, as mentioned above.

0xCCBA696 said: That would just move danbooru closer to being a pixiv mirror, tbh. I don't see this as needing to be something expressed outside of comments.

As if it didn't become one like a year ago anyway. I don't like that fact either but pretending it's not the case is pretty silly. Just about everyone is on pixiv now, and I've seen literally dozens of artists I used to follow move all, or most, or the complete/best versions, of their work to pixiv. I twitch every time I try to upload something brand new on an artist's homepage and and find out a better version was already uploaded a day earlier from their pixiv account I never knew about, but what can you do?

But anyway, if people don't think it's a good idea I don't feel strongly enough to argue further in favor of it either, so that's fine.

I'm with ultima on using the parent/child system for those kinds of images. I didn't quite understand the reasoning against it, I mean the line art is the same so it's the same image at it's core. You can even usually find all the variants using the similar search.

Even if I support that though, I don't think that has to rule out having the tag to identify them as an image response. Just parenting would show what the image is in response to.

Such a tag I think would be more useful for the images that have a few image response variants than the very large and well populated projects. The projects are likely to make it too hard for any of the smaller sets to be found without the use of a lot -pools in the search.

I agree also with using Parent/Child for this even if it seems like a stretch of the intended purpose of that function.

The only good way to make it more "kosher", would be to allow for multiple relationship types between posts (i.e. "parent/child", "image/response", "previousInSequence / nextInSequence", etc.).

This would involve modifying the system as a whole and adding new DB tables however, which I'm not sure Albert would want to do.

Parent/child is ideal for derivative images.

However, the pixiv parent is usually missing from danbooru (lineart and blank templates are less interesting, so they don't get posted).

Tagging collaboration (or colored, or colored_lineart) should solve the "two artists?" confusion. Pools are good for filled-in templates.

Some image responses borrow an idea/character/outfit from the parent, but it's not the same picture at all.

OK, forget about the pixiv clone argument.

How about this: under what circumstances would you be searching for images which were image responses? Other than administrative tasks, such as those which occasion searches for "parent:none" or "source:*fc2*" or whatever. What would be gained by putting that information in a tag rather than in comments?

I still oppose the use of parent/child for anything other than dupes or slightly modified versions, but it seems that many people disagree with me on that point.

I think the biggest argument towards structuring relationships is for machine-readability for future enhancements / third party systems to make use of the them. You can't get that via unstructured comments.

Right, but is "X image was a response on pixiv to another image on pixiv [which exists on danbooru as Y]" a relationship we want to be integrated into our database? I don't find it particularly germane. If we note such relationships which exist primarily on a level abstracted from the image itself, it seems to me that we are stepping onto a slippery slope in some sense.

However, if we just create a tag image response to note "X image was a response on pixiv to another image on pixiv" without noting what other image it is was response to, it seems pointless to me. What possible use could that be? Specifically, what would making such information machine-readable accomplish?

I agree with you on the latter, if not the former. Simply tagging image response is not useful.

I took the wording from pixiv (and because it's similar to wording elsewhere), but didn't mean to imply we would only be allowed to use it on images from pixiv. The only reason the whole discussion revolves around pixiv is because everything on danbooru comes from there lately, and it's the easiest way for artists to "respond" to each other in a centralized location. Without knowing Japanese and being familiar with two artists' personal websites, it's harder to tell when aritsts are responding artistically to each other. It'd be nice to capture that if we knew... though as others stated (even though you're against it), parent/child covers that already.

So again, I don't much care anymore if we use this or not. Parenting is fine I guess.

Well, instead of "parent-child" relationships, maybe we could develop something like "associated" images, without a strict hierarchy.

Are we drifting towards AdvancedRelationships ? Is this some sort of eventual necessity for any database of artistic works? :P (I'd be all for it, though I don't see a whole lot of applications on danbooru)

Well, it does seem like there is a call for some other sort of relationship grouping.

We have ways of dealing with large groups of images, both ordered and unordered (pools) that, vandalism aside, works pretty well.

We have a way of dealing with image variations (parent-child) that also works well.

But do those really cover every sort of way images can be associated? And if not, is there a call for a new way of doing it?

I would just like to point out that we've been on this slope for years now.

I see it as the inherent limitations of Danbooru's semantic model becoming apparent with its trade-off in usability and user-oriented utility over extensibility and future-proof design.

I also think that we have, by far, the better deal.

It does seem like we need some other ways of arranging images. The limitation of parent/child is that the parent is implied to be a "better" image than the child, but that's not always the case. The parent/pool/something else issue also came up in forum #24860.

I think collaboration makes more sense than image response when it comes to the pictures you're talking about (I suppose in part because I've come to automatically associate 'image response' to the expression charts and such.) I don't personally think of the parent as automatically being better, but I can see how that could be assumed. Quite a sticky situation.

But, as far as the image response charts go (like the wonderful hairstyles of my wife one, for example) ... I know some of them already have pools, but others don't. I was planning to create pools for the remaining ones that had multiple responses and was thinking that once all of the sets were pooled, we could make a directory to all of them either on its own wiki page or on the chart page.

...Would that be an alright course of action? I thought that having a place that had all the different sets listed would be convenient for people who enjoy them and upload them. (Well, I know it'd make it easier for me to find them, at least.)