Fat art database

Here is an idea that came to my mind a while ago:
A search engine where you can search for fat arts filtering by certain criteria:

  • Character name
  • Gender
  • Series name (anime, game, movie, …)
  • Body type
  • Approx. body weight
  • Artist
  • Scene: eating, sitting, exercising, posing
  • Date (when art was created)
  • Clothing (skirt, shirt, dress, …)
  • Hair color
  • Overall color tone of the picture
  • Possible other criteria I haven’t thought of yet
  • Species: human, centaur, snake, …

So that the user can search for various queries. For example:

  • Show all pear-shaped characters wearing dress, tone=yellow
  • Show all male characters from Full Metal Alchemist drawn in 2010…2015
  • Show all redheads under 200 pounds, sorted by newest

To build such a database we can either enter properties by hand (allowing everybody to add entries), or we can train AI to determine some of the properties. Things like overall color tone can be deduced programmatically.

What do you think?
I would not store the images themselves and store only direct links to save storage costs.
Also I am a web developer, not so much a game developer, so programming such a thing would fit my area of expertise well; it seemed like a good idea for a project that I myself would work on

10 Likes

While I think it would be a very good resource, it would require quite a bit of work. A similar idea come from image booru such as https://grognard.booru.org/ (where I get a lot of character art for RPGs), which have users upload and tag photos. One issue comes with having a large enough base of people willing to fully tag images. Perhaps having a very curated list of tags like you made might help with that.

I hope you find some support to make this because it would be very helpful when trying to look for specific art/artists.

3 Likes

That’d be interesting, if someone had the resources and the know-how. I’d love to see a site like that! But in the mean time, what exists works for me. Allow me to reiterate: This would be great, but feels a bit like a pipe dream at the moment. If only…

2 Likes

Sounds like a lot of work, but it’d be nice to look at such cuties without any… y’know… less cuties.

2 Likes

Well you might be in luck considering how Eclipse on Devianart is complicating matters on several levels.

Whether some are willing to jump ship is something I can’t say but for the most part some have left thanks to Eclipse.

2 Likes

If you’re not looking to store the images then Tagpacker might be a good program to use or get some ideas from

1 Like

Hoo boy, definitely a bad idea, I have to be honest. The internet’s much too big of a place to just shotgun image searches across it and tagging everything like that. I can imagine MAYBE doing that selectively, like building a browser plugin for a few art sites on the internet, but that’s just a waste of time.

My suggestion is to say fuck overthinking it, fuck months/years of coding, fuck assuming everyone has the same concept of who’s “obese” and who’s an “SSBBW” and just pick an existing booru, then getting more people to join THAT, then submit content and tag things appropriately there. I’m sure there’s at least 20 different fat-centric boorus out there. Just find one and help draw attention to it.
Booru galleries already have a solid tagging system and the means to search it.

1 Like

Also I don’t like the idea of an AI trying to tag all these things automatically. An AI like that would have to be awfully powerful, trained on a massive model that covers every variation of everything ever, including multi-dicked, blue-haired herm furries, etc.

I dunno… programming a half-decent game seems like an immeasurably greater effort than programming a few web pages running simple DB queries

honestly, this seems like a pretty cool idea. i’d love to be able to find specific types of fat art (like sfw fat dudes with a certain body type) without having to see everything else

also a suggestion for a criteria you can search for, nsfw or sfw. i know a lot of people (like myself) don’t want to see nsfw stuff for this.

2 Likes

I’ve actually been working on such a thing very indirectly. I have an absurd amount of fat art stored, over 4000 images across a slew of devices but I’m not sure on the exacts, but even still I think it’s a lot. While this was originally me just engaging in epic degeneracy, I realize this could actually be a good start to making such a database. The only thing we need is a place to store it

1 Like

I think I have that beat by at least an order of magnitude! And as much again of fat fur artwork.

The problem with this idea isn’t doing it technically, it’s doing it at scale. Let’s say it takes me 5 mins to look at a piece of art, tag it with all the fields suggested in the OP, and upload it (sorry, I can’t link to it, because much of this stuff came from Usenet, and the originals are gone, or it was on the web but the artist had a clear out, or deleted their account … which is afterall, the reason for saving it in the first place). That would amount to about 5 man-years of work to classify it all. Now, colour me stupid, but I don’t think anyone is about to pay me a salary for five years to curate my art stash! Just de-dupping it takes hours of CPU time.

But, I’m sure lots of other people have saved the same art I have, so the system is going to have to be able to detect duplicates, and say to the users “I’ve seen that one, and you don’t need to tag it”. And it will need to look at the images, because the same art will be on DA, FA, Twitter, Weasyl, InkBunny, and countless chans, boorus, and torrents - the URLs won’t match, the filenames won’t match, it’s been altered, edited, resized, flipped, metadata changed, you name it. It’s not going to run on a quarter blade on cheap hosting.

And god forbid it works, becomes popular, and gets a lot of traffic. Scaling a database across multiple blades to support many many parallel queries on an ever growing dataset is a non-trivial task.

2 Likes

What if a lot of people work on a project like this?

2 Likes

I personally think this is a really convenient idea if people can pull this off, this can drastically speed up people looking up art if done properly.

However I might throw in my two cents in to add some more, if however this does get achieved it might be really easy to then pivot the data base to be used for machine learning for creating machine made drawings (Like for example MakeGirlsMoe - Create Anime Characters with A.I.!), which could be a great asset for people who need custom characters/ larger characters drawings without being forced to draw the character or pay for a commission of several characters. That could drastically improve the quality rpgmaker games as creators can use ready made assets that aren’t solely in the creator.

Although I believe there’s several issues that may have to be addressed to be able to realistically accomplish the database

  1. Making a system in the database that recognizes repeating images of the base image and deletes them, although having a user upload a higher resolution version of the original in the database would cause the database to use the new resolution version as the base image. This can allow multiple users to upload images without the issue of causing massive influxes of repeat images. This can help cut down on the need for moderation.

  2. Having a clearly defined set amount of filters so that each image can be defined in the correct tags. This mainly just entails about the OP said about criteria tags, however it might be worthwhile to add in tags for the artstyle aswell to be included with the direction of character, poses/action character is doing, and amount of character shown in the shot (is it just a bust drawing, or a full figured drawing?)

It also would be difficult to have a general consensus of agreed upon body weight for characters due to artstyles making it difficult to gauge how much a character weights. A concept I might put forth instead of a weight system is more of a tier system of a progression of a character’s encumbrance (for example tier 1 being normal bmi, tier 2 slightly chubby with beefed up parts, 5 defined pot belly with flaired assets, 10 bbw, 15 ssbbw with a defined split stomach down to their knees, and rear takes up two chairs, etc, etc) This would likely have to be done as a progression drawing all the way to absurd sizes in a sense similar to bra sizes, that may be the most straightforward way to have a defined measurement of a persons assets if I’m not over analyzing it. Having a clear tier of assets at different sizes can also be easier to define what size each asset is compared to the body, so instead of looking for a pear shaped character instead of pear shaped it would more be like find character with tier 10 and above tush and tier 8 and below for every other body asset.

If I’m to be harsh, there might also have to be a filter for quality of work too, there’s a clear line between a person who uses the circle tool in mspaint and a cave dweller using professional equipment and years of experience of drawing degenerate fatty wank for coom cash. This could be a pretty controversial tag, so it might be best to make it extremely simplistic as of having a 1-4 range. with 1 being mspaint stuff, 2 being better than mspaint but not good enough to be decent ,3 being decent, and 4 being professional work. The quality tag could be cemented better with clear definitions as to what makes a good character art, be it with consistent adherence to some form of anatomy, lighting/shading, clean lineart, etc etc. While not needed it could help filter out tags to a greater degree.

I mean if you’re making a database with filters you may aswell have the database with the most filteriest filters, that badlandschugs can use it as a lifestraw to drain the ocean in record time.

  1. The possibility of changing those filters due to human error; there’s a definite likely hood of users mistakenly inputting in the wrong filters when submitting an image, this could be circumvented by adding in a flagging feature that allows users to suggest to admins or mods to change the filters to best suit the image. Or having the community change the tags themselves by putting in enough flags to change the tags.

I think currently the best line of action to handle this is to;

  1. plan specifically what guidelines to implement into the system
  2. produce a system to recognize repeating images to delete and make a higher resolution image the base image
  3. implementing the image tagging system,
  4. making a flagging system to correct any errors in tags (making it able to be managed by the community would make it more autonomous and less of a pain for admins)

I don’t think it’ll be as big of a hassle as it would be to make a AAA game, but it would likely be a very sizeable upbringing to bring into fruitition that’ll require a fair amount of planning to make happen, and more problems would arise after the development when people submit images than a game would make.

Apologies if I rambled through this or it was worded poorly, I wrote this when I was tired and I didn’t get my chicken tendies today.

3 Likes

I would be so down for contributing to something like this. However, I would need some guidance on what to categorize things into albums and such

Maybe with this, we will be able to complete the world’s most legendary and mythological task: Finding all of the pages for Layla’s Stuffed Adventures

2 Likes