Unsourced + Modified posts

First concern is resolved.

Then my second concern is that there are posts like post #2687625 that come unsourced, and when doing a visual comparison with the original image (sourced from the DL edition of COMIC Penguin Celeb 2017-03), it turns out that the saturation levels on both are different. Doing a pixel-by-pixel diff with magick results with this. When searching for the image uploaded to booru (with the changed saturation levels), it turns out the image is only indexed here, and there aren't really any other signs on other websites. Searching for the original, unmodified DL edition image results in a much larger amount of hits, mostly from other Japanese websites that can confirm the validity.

I'm not really a particular fan of third-party modifications if they are of blatantly digital sources (as post #2687625 isn't derived from a scan) but others might think differently. post #2682421 is from a scan and the saturation is also turned up from the original source, so it's somewhat acceptable, albeit I'm especially leery on approving things like those.

So then the second question is: To what extent are images that are modifications of images from a digital medium also allowed? The parent of post #2354256 is quite literally a valid fix/improvement (although it could be done any infinite number of 'valid' ways, theoretically), but from an intentionally subjective point of view, would post #2687625 count as an 'improvement' -- and would we have to consider other continuous 'improvements' like those valid for approval? Again, theoretically there are "infinite ways" to improve an image, so picking just one or two to keep is particularly sketchy. We aren't primarily focused on hosting third-party edits, after all.

Perhaps this is more concerning that particular user because he has a habit of doing this on nearly every upload he has unsourced, which I find somewhat disdainful.

Pinging @CodeKyuubi and @☆♪ for opinion on the matter.

EDIT: Accidentally words

I would consider nhentai an inferior source to the source e-h sites, and because of the nature of being resized, I'd be fine with nipping it in the bud while it's still not a common source and consider it a source of image samples, unless where other versions cannot be found (Much like with sankaku). Still, even if it doesn't exist on the main e-h galleries, it does not mean that the torrents for those galleries are dead, and the images can still be found in the numerous torrent packages distributed on ecchi-dl sites.

From what I can see, post #2687625 in question is oversaturated (Exposure is a different beast), because on my color-calibrated monitor the original looks fine. This is an issue with monitors that aren't calibrated, as colors can appear differently or subdued. I understand their desire to provide an 'ideal' image for the end-user, but I can also see that it can be a snowballing issue that can turn out much the same way gorua_(youce01)'s many revisions bloat his page. I personally used to light balance digital 'scans'. One thing to note is that official digital distributors such as melonbooks and pixiv booth often do not release true-digital images in their inventory, but just well-scanned scans (And often with cropped edges that do not seam well together in multi-page spreads).

But back to the topic, at the moment I think it's fine if we accept photo-edits in good faith, until users try to abuse it to gain upload count or mess with parent/child relations. Understandably, this can also coincide with the touchy subject of artist intent. For example: post #1860988 vs post #2661855. Though a digital source, after some consideration I ultimately decided to color balance the image to remove what I felt was overly-excess red, even assuming a sunset light source, something hardliners might find unacceptable.

https://puu.sh/viwYk/dd5ac99710.jpg
I don't think it's wrong to mess with the original (actual) scan in endeavoring to improve it (See above link). I've cut out a part of the near-final product I've got going to show the source scan underneath, while the top has been balanced to be as close to the artist's version as possible.

I can tell you that the mindset behind not sourcing when altering the image is that, because it has been altered and the md5s are different, the user believes that there is no proper source for it anymore (I used to do this). What should happen is that he should be informed and educated that the used source should be linked, either via the source bar or by comment (Where md5 mismatch tags might be avoided. post #2596521 for example, its true source is imgur linked directly by the artist through plurk (Some sort of Chinese facebook), but some misguided soul replaced it with the twitter link, which also triggered a bot response).

I'll abide by the stance that *most* images from nhentai are image samples and should probably be replaced (some are originals, strangely), albeit as of right now they are of very low priority to me (given the availability of the original scans). In addition, currently there's a new feature being worked on to enhance sample replacement without having to resort to a post deletion (favorites' ordering and score are still intact, etc). My main priority on that front will still be replacing twitter samples.

Regarding post #2687625, that's what I was thinking -- the original also looks more than acceptable on my screen; publishing companies usually do a good job of providing decent quality digital editions of their print magazines, so there's next to no need to modify it IMO. Seems like a really unnecessary thing to do.

I take a harder stance on third-party edits with digital art more than before, but I think it's important we at least outline some guidelines about what makes a truly acceptable third-party edit versus just uploading the original (if it wasn't already available before). See, in cases like post #2354256 and its' parent, clearly the original would be unwanted (as the publisher likely modified it to prevent distribution), but in cases where the original is preferred inline with any available third-party edits it would be good to know as approvers the things to keep in mind when approving such posts. Something like post #2684763 is completely fine for me, but edits that make wholly unnecessary modifications (save for good decensoring as say, post #413569) should be completely disregarded. That is, things like nude filters, poor decensoring (post #2333202), and adding of unoriginal features (post #2585351).

I'd like to hear more thoughts on this though, since I usually don't like it when there aren't clear guidelines to abide by. Sure, I can continue to approve things I think are acceptable assuming good faith, but the whole modification jive that some users are having sort of diminishes that ethic; to the point where I rarely approve images that are unsourced or are from non-trustworthy websites, like random .ru websites. It's not healthy for aspiring uploaders either, knowing that other users have that avenue of 'cheating' their way to a high post count if they don't bother to source and/or legitimize their posts as a loophole.

Mikaeri said:

I'd like to hear more thoughts on this though, since I usually don't like it when there aren't clear guidelines to abide by.

I wasn’t sure whether I should post my thoughts on it, but as you explicitly asked for input, here are my critical thoughts on it.

I’d roughly distinguish between three cases: technical adjustments, alterations and restorations.

A technical adjustment for me is whatever is necessary to compensate some unavoidable technical shortcoming, like white-balancing a scan.

CodeKyuubi said:

https://puu.sh/viwYk/dd5ac99710.jpg
I don't think it's wrong to mess with the original (actual) scan in endeavoring to improve it (See above link). I've cut out a part of the near-final product I've got going to show the source scan underneath, while the top has been balanced to be as close to the artist's version as possible.

This is perfectly fine in my opinion. Any scanner invariably modifies an image by producing a scan that doesn’t match the source material. The person making the scan should try to undo those modifications to bring the scanned image as close as possible to the original, which involves having the original to actually look at it. From what I can tell, CodeKyuubi seems to be doing a good job at it. Simply taking whatever the scanner spits out, possibly even with some magical auto-correction, would be worse.

An alteration for me is anything that was altered from how the original artist intended it. The easiest example are of course nude filters, but this would also include brightening an image just because someone thinks it’s “too dark and depressing”, which might just what the artist had intended. Anything like that is just right out.

Now restorations is a big gray area that covers everything where one modifies an image in order to restore the original artist’s intention, such as redrawing covered or missing areas (de-texting, de-censoring, joining a double-page spread with the middle missing), but maybe also adjusting colors and levels of an image that seems off. The problem here is that one has to guess what the artist’s intention was and the guess can be wrong. That’s important to me because I highly value the artist’s intention and original work. The larger the redrawn/modified area, the more likely it is that one messes up, besmirching the artist’s work.

I’ve been doing redrawing for scanlations for a while and I always struggle with guessing what the artist originally had there. How thick and even does the artist draw lines? How does the artist draw hair? Hands? Place screentone? What kind of object might be placed there? A few days ago I made a small redraw for post #2685152 and even though it was a tiny, simple one, I had no idea what the short line above the arrowhead was for. It wasn’t on the other foot and the super-deformed characters in the artist’s other posts didn’t have it either. It was undoubtedly there, but what was it for? My guess was as good as anyone else’s, which is exactly why I think that restorations should be avoided if possible. The same goes for color-adjustments, even if done in good faith. Without having the original, how would one know what the digital image should look like?

The reason why I made that small redraw after all was because when the characters are actually printed and cut out, most of the arrow is removed and leaving part of the arrowhead there looks really stupid. But as it was an unnecessary third-party edit, I didn’t upload the modified image to Danbooru. I didn’t put the whole modified image elsewhere because I don’t want to upload another artist’s work to a site I cannot influence. I hope that even MS Paint will be enough to put that little patch in place for anyone who wants to print it.

Mikaeri said:

Again, theoretically there are "infinite ways" to improve an image, so picking just one or two to keep is particularly sketchy. We aren't primarily focused on hosting third-party edits, after all.

This is pretty much the gist of it.

CodeKyuubi said:

Understandably, this can also coincide with the touchy subject of artist intent. For example: post #1860988 vs post #2661855. Though a digital source, after some consideration I ultimately decided to color balance the image to remove what I felt was overly-excess red, even assuming a sunset light source, something hardliners might find unacceptable.

I have to admit that I haven’t actually looked at it because I have some of the tags blacklisted, but from what you described, it’s an edit that I don’t approve of. If the majority of viewers agreed that there was too much red in the image, then it might’ve been a poor choice by the artist, but that still doesn’t mean that you should tamper with it to “fix” it. Maybe the artist really liked it the red way and you changed the image in a way that the artist wouldn’t like.

Mikaeri said:

Something like post #2684763 is completely fine for me, […]

Eh, that’s quite a lot of magic that happened there...

Sorry for rambling a bit in the middle. Hopefully, it’s at least a bit helpful.

kittey said:

From what I can read, I can't tell a difference between your perception of alteration and restoration in regards to color/light balancing.

What I can tell you is that, setting aside artist intent, there is ideal white and color balancing based on technology to detect upper and lower bounds of exposure and color information and what have you. Whenever I work with images with no direct digital comparison, that is what I aim to achieve. The process is the same as with actual photo-editing, where your end goal is to have a color/light-neutral picture.

Do you suppose we should have some sort of help:third-party edit page to elaborate on these kinds of cases?

As for post #2661855 (original) I think it's fairly debatable... Albeit it is an arguably better edit in terms of color contrast, it still feels somewhat off from what the artist originally intended. The original is likely purposely red/brownish given the comparison with the child posts.

Just from reading though, the restoration part is really difficult to determine... One can say that all technical adjustments are restorations (this is especially true of scans), but not all restorations are acceptable if they stray too far from what the artist intended. In those cases it would be considered an alteration.

I'll just try to outline concisely what I think here.

• Most high-quality technical adjustments (especially of scans and scan restoration, or of intentionally modified digital distributions by the original publisher) are acceptable to approve. Things like removing scan artifacts, fixing bad stitches, etc.
• Some restorations are acceptable. These are things like third-party decensors, color balancing, and redraws. This usually assumes good faith in what the original artist/publisher intended to distribute.
• No alterations are acceptable, especially considering they are third-party modifications that add unoriginal details to the image: nude filters, adding a pussy when no pussy is drawn, etc.

I don't have much to say that hasn't been said, but since I was summoned here goes:

Digital and scans should be treated separately, and a scan I'd consider pretty much anything that has been through an analog stage at any point. For example post #2354256 is likely a "scan" even if it came from an official site of some sort.

Once there's a conversion from analog to digital, there's no longer the concept of perfect preservation that you have with a pure digital source. The scanner's output is dependent on lots of things and isn't more faithful to the original than a color-corrected version. So correcting scans to fix colors, fix stitching, remove staples, etc. is desirable in my book - in fact, I'd consider it part of the scanning process. However it isn't a simple task (I wouldn't trust myself to do it, despite my knowledge of the digital side of things). CodeKyuubi knows what he's doing, but clearly not everyone does, so "fixed" scan posts should be subject to more scrutiny to make sure it's not just someone making it look good on their monitor and throwing away actual color information. Anyway, there seems to be at least a general consensus here on how to treat scan modifications and I pretty much agree with it. I might just add that if the modifications are significantly subjective, it may be worth uploading the original as well, as a child. It doesn't cost us a whole lot to have the extra post here.

When I saw Mikaeri's suggestion for a third-party edits wiki, my first reaction was that some of that was already on howto:upload, but after checking there I see that it's not mentioned at all. We should have something on that page about editing scans, and the distinction between digital and analog sources.

We should probably take a harder stance on unsourced uploads. A lot of these issues come up when we don't know where an upload actually came from. Uploads with no source whatsoever are very suspicious. If you scanned something yourself you can just put the name of the publication you scanned it from in there. If there is an original image file URL but you don't want to use it as a source because you made some legitimate modification, you should at least leave a no-bump comment explaining the situation and the closest source.

Some random ideas:

• Uploads with no source could be distinguished somehow in the mod queue, and would be subject to more suspicion there. This would give approvers a chance to head off both accidental and intentional source omission early, without adding yet another thing they'd have to remember to check for. This could potentially also be extended to mark posts sourced from sites known for sampling, etc.
• Some kind of warning if you try to upload with the source field empty. It shouldn't absolutely stop you from uploading, though, as there are some cases where there really isn't anything reasonable to put there. (Usually a no-bump comment would be warranted in those cases.) I'm not talking about another bold message at the top, by the way, there are already too many of those. I mean something that bothers you when you actually try to submit.

Is there any legitimate, common case where no source whatsoever is warranted?

Yeah, it's not mentioned on howto:upload at all for what third-party edits are and aren't desired. As for why I think a wholly separate linked wiki page would do better is because howto:upload is already inundated with LOTS of information about acceptable content under the guidelines. Since third-party edits are of a complicated nature whether to keep or delete, it's better to keep it separate and elaborate when needed, and just have it linked in howto:upload.

For approvers I think it's especially important to note since it's easy to make a mistake about what seems genuinely acceptable. post #2679645 for example. The original scan that that post was derived from was not linked by the original uploader, and neither was it properly tagged with the relevant modifications applied.

Not only that, but the criteria for which third-party edits is also under subjective scrutiny. And where the original uploader did actually make modifications, it might actually be better to treat them as md5 mismatches or image samples and just replace them entirely with the original if it wasn't uploaded yet. The upcoming feature might help with that.

I think it's better to take a hardline stance towards unsourced uploads, and I would consider highlighting them in the queue with red or yellow since there are uploaders out there that make it a constant habit to not source their uploads when absolutely necessary. Unsourced game cg or the like is somewhat okay (since a rip can be easily found) but for other types of posts, not having the source should really be detriment to a post's approval. Luckily more are picking up on that, but some aren't.

☆♪ said:

• Uploads with no source could be distinguished somehow in the mod queue, and would be subject to more suspicion there.
• Some kind of warning if you try to upload with the source field empty.

These are both good ideas. I'll see about implementing them.

Mikaeri said:

Yeah, it's not mentioned on howto:upload at all for what third-party edits are and aren't desired. As for why I think a wholly separate linked wiki page would do better is because howto:upload is already inundated with LOTS of information about acceptable content under the guidelines. Since third-party edits are of a complicated nature whether to keep or delete, it's better to keep it separate and elaborate when needed, and just have it linked in howto:upload.

I pushed for this seven years ago in topic #3568 to no avail. It's next to impossible to get approvers to agree on even the simplest guidelines. At least it was back then, maybe things have improved since then with newer generations of approvers.

If there's no disagreement over it, I'll go ahead and make the change to include that third-party edits are allowed under very strict circumstances. I'll try to include the ideas that everyone has listed here when creating it.

EDIT: Although perhaps having a tag in the queue can help with these kinds of things. I imagine it'd be useful if a third-party edit tag would be a sign to treat an unsourced post differently. Having the uploader being required to comment information about that would also be useful.

EDIT 2: Finished

Mikaeri said:

If there's no disagreement over it, I'll go ahead and make the change to include that third-party edits are allowed under very strict circumstances. I'll try to include the ideas that everyone has listed here when creating it.

EDIT: Although perhaps having a tag in the queue can help with these kinds of things. I imagine it'd be useful if a third-party edit tag would be a sign to treat an unsourced post differently. Having the uploader being required to comment information about that would also be useful.

EDIT 2: Finished

Would images such as these fall under that tag (user:chinatsu_(sweetpea) source:*youtube*)? I screencapped them from artist videos and cropped the images, saving them to PNG. I had made a thread about this some time ago but didn't have an appropriate tag for them.

chinatsu_(sweetpea) said:

Would images such as these fall under that tag (user:chinatsu_(sweetpea) source:*youtube*)? I screencapped them from artist videos and cropped the images, saving them to PNG. I had made a thread about this some time ago but didn't have an appropriate tag for them.

I would say no, but they should be tagged somehow. It's a bit of a grey area: Assuming the screencap is done right, it's a perfect digital copy of what was in the video, so it's not really edited by a third party. However, the video itself is lossily compressed and the image that was on their screen while they were recording or that they edited into the video is quite likely scaled, so it's not a perfect copy of the original image. If a more direct source showed up, it would (have the potential to) be superior.

How are you making those, though? Are you just taking a screenshot of the player in the browser? I downloaded the video for post #2587370, took a frame from around 00:09:09 and cropped it (leaving off a few pixels at each edge that were blended into black), and got this, which is noticeably different from yours. Did you get your screenshot from a different place in the video? If not, it's scaled and possibly otherwise processed.

☆♪ said:

I would say no, but they should be tagged somehow. It's a bit of a grey area: Assuming the screencap is done right, it's a perfect digital copy of what was in the video, so it's not really edited by a third party. However, the video itself is lossily compressed and the image that was on their screen while they were recording or that they edited into the video is quite likely scaled, so it's not a perfect copy of the original image. If a more direct source showed up, it would (have the potential to) be superior.

How are you making those, though? Are you just taking a screenshot of the player in the browser? I downloaded the video for post #2587370, took a frame from around 00:09:09 and cropped it (leaving off a few pixels at each edge that were blended into black), and got this, which is noticeably different from yours. Did you get your screenshot from a different place in the video? If not, it's scaled and possibly otherwise processed.

Here's the thread I created some months ago: topic #13170

My method for all the screencaps I made is the following:

• Download highest resolution using YouTube-DL
• Open video in mpv and use itss screenshot keybind and the appropriate time in the video
• Open in GIMP, crop out the image, export to PNG to prevent further data loss

I considered doing a more precise screenshot with say FFmpeg or something else but had some difficulty in setting that up for Windows.

I've tried to do things in a way to ensure the best quality despite the problems you pointed out in the reliability of the artist's video. And sure - doubtlessly if they uploaded the image elsewhere that would be the preferred version - but I did this for cases were that was not available.

Not sure how I ended up cropping part of the girl's head out, but well it happened. I don't object in the slightest to you uploading your crop and flagging mine if you'd like to.

chinatsu_(sweetpea) said:

Here's the thread I created some months ago: topic #13170

My method for all the screencaps I made is the following:

• Download highest resolution using YouTube-DL
• Open video in mpv and use itss screenshot keybind and the appropriate time in the video
• Open in GIMP, crop out the image, export to PNG to prevent further data loss

I considered doing a more precise screenshot with say FFmpeg or something else but had some difficulty in setting that up for Windows.

I've tried to do things in a way to ensure the best quality despite the problems you pointed out in the reliability of the artist's video. And sure - doubtlessly if they uploaded the image elsewhere that would be the preferred version - but I did this for cases were that was not available.

Not sure how I ended up cropping part of the girl's head out, but well it happened. I don't object in the slightest to you uploading your crop and flagging mine if you'd like to.

That's very interesting, because I used almost exactly the same process. Just used a different program to crop, but gimp doesn't do anything weird with cropping AFAIK, so I can't figure out why your crop ended up still being 1080 pixels high. It's like you had a pan/scan video filter on in mpv or something. I just tried post #2484264 and I get the same thing you did (I just cropped off one extra pixel on the right), so whatever happened with that D.Va doesn't look like it's a consistent issue with your setup, at least. I considered that a different version of the video might have been uploaded when you took the screenshot, but I don't think that's very likely. The only other thing I can think of is that it's related to Windows and the way mpv uses its graphics stack somehow - I'm on Linux, which mpv supports better. I guess I'll go ahead and upload mine, and we can hope it was a one-time fluke in your process. It sounds like the way you're doing it is solid overall.

EDIT: Actually, one note for next time: Whatever encoder you're (gimp is) using is using a very low compression level. Running optipng on your file cuts its size almost in half. In the future, you should turn up the compression level when you're exporting the PNG - it's still lossless, it just spends more time trying to make a smaller file.

Anyway, I'll go bump your other thread about a new tag so we don't derail this one. I wouldn't consider these third-party edits for the purpose of this discussion. If done correctly, they're unmodified semantically from the artist, just undesirably processed by a hosting service - similar to how Twitter compresses jpegs to hell.

For what it's worth, photoshop PNGs are very optimized. I've tried optipng'ing a few PS PNGs and they either removed a literal handful of bytes or returned an error that no optimization could be made.

CodeKyuubi said:

For what it's worth, photoshop PNGs are very optimized. I've tried optipng'ing a few PS PNGs and they either removed a literal handful of bytes or returned an error that no optimization could be made.

GIMP does fine too, if you turn the compression level up to 9. Maybe not as good as photoshop, but to the point where optipng trims single digit percentages at most.

Hmm, I suppose I should optimize some of my end card screencaps sometime then. I took them straight from the player, so they're quite bloated.

☆♪ said:

That's very interesting, because I used almost exactly the same process. Just used a different program to crop, but gimp doesn't do anything weird with cropping AFAIK, so I can't figure out why your crop ended up still being 1080 pixels high. It's like you had a pan/scan video filter on in mpv or something. I just tried post #2484264 and I get the same thing you did (I just cropped off one extra pixel on the right), so whatever happened with that D.Va doesn't look like it's a consistent issue with your setup, at least. I considered that a different version of the video might have been uploaded when you took the screenshot, but I don't think that's very likely. The only other thing I can think of is that it's related to Windows and the way mpv uses its graphics stack somehow - I'm on Linux, which mpv supports better. I guess I'll go ahead and upload mine, and we can hope it was a one-time fluke in your process. It sounds like the way you're doing it is solid overall.

EDIT: Actually, one note for next time: Whatever encoder you're (gimp is) using is using a very low compression level. Running optipng on your file cuts its size almost in half. In the future, you should turn up the compression level when you're exporting the PNG - it's still lossless, it just spends more time trying to make a smaller file.

Anyway, I'll go bump your other thread about a new tag so we don't derail this one. I wouldn't consider these third-party edits for the purpose of this discussion. If done correctly, they're unmodified semantically from the artist, just undesirably processed by a hosting service - similar to how Twitter compresses jpegs to hell.

Regarding PNGs, I recall turning up the compression. I supposed I should have optimized using optipng after exporting.

And I made the dva screencap months ago. I think I just cut too much off the top and didn't double-check. Thanks for uploading your version.