掲示板 Forums - 2026 Kao Coin Contest - Rule preparation
Top > renshuu.org > Announcements
Page: 3 of 3
Getting the posts
Page: 3 of 3
Top > renshuu.org > Announcements
Requirements for all submissions:
1. The final Kao-coin image.
2. The kao-coin suggested name (in English and Japanese).
3. At least two sequential workflow proofs (of any planning/development/progress stages) that logically, clearly, and directly align with the final submitted image.
4. A short description of the materials, tools, techniques, and references used for the work.
5. An agreement stating that no part of the work (from idea generation to editing to formatting, etc.) was done with the use of Al, and which could be a checkbox to be marked before sending in each submission (so as to nix any possibility of alleging ignorance on the matter).
Additional requirements only for entries selected by Michael as potential finalists/winners (in order to reduce the task of having to do this for every submission), especially if flagged through the "Report Content” link:
6. Any number of additional sequential workflow proofs like those provided in requirement 3.
7. Visual proof of a random sampling (chosen at Michael's discretion) of the materials, tools, techniques, and/or references used to create the work as described in requirement 4.
8. A high-resolution (illustration or photo) file in a lossless format such as TIFF, PNG, or uncompressed JPEG.
9. The original work files (layered illustration or unedited photo) with complete metadata—for automatic elimination of any entries with conflicting, lacking or incomplete information—to compare with requirement 8.
All participants can prepare requirements 6–9 in advance in case they are selected as potential finalists/winners to avoid not being able to meet these requirements if asked to do so.
I feel like that's all a bit overkill and would just discourage a lot of genuine artists from participating at all. This isn't a final exam for an art class. Honestly I feel like requirement #9 would be sufficient enough as proof for a finalist, even a combination of 8 and 9 would be good.
I think a video of the drawing process should be part of the submission. It's not hard for someone working physically to set up a camera at an angle. And AI generated videos are much harder to pass as real than images. Additionally, perhaps add something in the video that would be hard to fake by AI, like scrolling through or showing a large piece of text, as letters usually get deformed in AI generated content. If there's some AI that can make a realistic video in this way and only a few poeple have access to it, I'd say to simply accept the possibility of a few AI generated images making their way, as long as one can demonstrate that it would be only a small part and no user will be able to tell. Maybe make the video proof public too and have a simple reporting system, in case Michael misses something while checking the video.
This might be similar to what ダニナド said but maybe
A. For physical works, video of the artist wobbling the paper around, spinning it around, flipping it back and forth, something like that seems like it'd be hard for AI to recreate (especially the flipping from back to front). Although I could see poor quality videos seeming like AI even if it isn't.
B. For digital works, assuming the user has multiple layers, screen recording (or a camera pointed at their screen) of them drawing a line through the layers in such a way an AI could not create. For example, lets say the shading and flat color are their own layers, if the artist made a layer between those two and drew some wavy lines through it (temporarily), it'd likely be difficult for an AI to recreate in any accurate way. Issue with this could be that some artists may not actually do layers which could lead to that verification process being impossible.
B2. Reordering the layers temporarily, like juggling them around. This might even work for the layer minimalist artists since I think even most artists choose to make at least 1 layer for lineart and 1 layer for color. Shoot, maybe even move the layers around a little bit individually.
I'm essentially saying we should get silly with it to prove its authenticity.
Requirements for all submissions:
1. The final Kao-coin image.
2. The kao-coin suggested name (in English and Japanese).
3. At least two sequential workflow proofs (of any planning/development/progress stages) that logically, clearly, and directly align with the final submitted image.
4. A short description of the materials, tools, techniques, and references used for the work.
5. An agreement stating that no part of the work (from idea generation to editing to formatting, etc.) was done with the use of Al, and which could be a checkbox to be marked before sending in each submission (so as to nix any possibility of alleging ignorance on the matter).
Additional requirements only for entries selected by Michael as potential finalists/winners (in order to reduce the task of having to do this for every submission), especially if flagged through the "Report Content” link:
6. Any number of additional sequential workflow proofs like those provided in requirement 3.
7. Visual proof of a random sampling (chosen at Michael's discretion) of the materials, tools, techniques, and/or references used to create the work as described in requirement 4.
8. A high-resolution (illustration or photo) file in a lossless format such as TIFF, PNG, or uncompressed JPEG.
9. The original work files (layered illustration or unedited photo) with complete metadata—for automatic elimination of any entries with conflicting, lacking or incomplete information—to compare with requirement 8.
All participants can prepare requirements 6–9 in advance in case they are selected as potential finalists/winners to avoid not being able to meet these requirements if asked to do so.
I feel like that's all a bit overkill and would just discourage a lot of genuine artists from participating at all. This isn't a final exam for an art class. Honestly I feel like requirement #9 would be sufficient enough as proof for a finalist, even a combination of 8 and 9 would be good.
I understand your point. Maybe I was not entirely clear. What I am suggesting is a combination of any of the points from 6 to 9, which is why I said “if asked to do so.” :)
However, I don't believe any of those requirements would be too much of a burden for anyone even if all of them were asked to be met since, after all, all that participants are being asked to do is to document different parts of their workflow and creative process as they are working.
Also, perhaps adding a bit of uncertainty to some of the possible requests (what points 6 and 7 are really about), might reduce potential AI submissions on account of the difficulty of having to create random proofs to try to match a non-existent workflow or creative process later on—especially for those that use AI because “it's easier,” as the idea is to make it troublesome for cheaters while keeping it manageable for everyone else.
And since not everyone works in the same way (some artists might not use references or might not work with layers) or with the same setup (some artists might not be able to record a time-lapse video while they work), I believe it's also important to have some flexibility through variety regarding these proofs precisely to avoid restricting participation. And having this variety also makes it easier to detect AI submissions, because although videos, metadata, and even some layered and lossless files can be faked well enough if you know how to do it, having several different types of supporting evidence that should consistently validate a single piece of work is much harder to fake with AI.
For me, having to provide extra proof for any submitted work is absolutely worth it if it means keeping AI slop out of the equation as much as possible and celebrating creativity through 100% human-drawn art.
In any case, I hope that a good set of rules can be created from all of the wonderful ideas posted here so that the Kao-coin contest can go on this year and for many more to come. :)
I think a video of the drawing process should be part of the submission. It's not hard for someone working physically to set up a camera at an angle. And AI generated videos are much harder to pass as real than images. Additionally, perhaps add something in the video that would be hard to fake by AI, like scrolling through or showing a large piece of text, as letters usually get deformed in AI generated content. If there's some AI that can make a realistic video in this way and only a few poeple have access to it, I'd say to simply accept the possibility of a few AI generated images making their way, as long as one can demonstrate that it would be only a small part and no user will be able to tell. Maybe make the video proof public too and have a simple reporting system, in case Michael misses something while checking the video.
Thank you all and for the renshuu.org team for contributing and hosting this event annually. My daughters always look forward to it, and I hate to see it ruined by AI-generated works.
I think recording a short video manipulating the work (physical paper or in their drawing app-of-choice) and stepping through the process or layers should be good enough. It would already be an extra burden for マイコー and team to review those. To be clear, the user's face wouldn't have to show up in the video, just a clear clip of the work/UI being manipulated by hand. I would think an AI could fake a convincing looking UI (to the untrained eye) but it would probably make mistakes that real users of Photoshop/Clip Studio/Krita/Sketchbook would notice. Adding hands and real ambient audio to the mix would hopefully make it more effort than it's worth for someone trying to game the system.
It's a bit sad that it has come to this. My family and I participate every year (with 1 silver award!) and we look forward to it again this year if it can be done without too much burden on マイコー and team!
Thank you again for all of your hard work! Would love to participate again without taking all the fun out of it!
note - to clarify - if a work is already drawn on paper (because user started early, or already finished), it may be difficult to show "progress", so manipulation of the paper and explaining the concept may have to be enough. It's difficult to cover all the cases...
I agree with others on timelapse as a potential option but with time-lapse getting easier for AI to fake, I think the hardest for them to emulate would be a straight up screen recording (think OBS, free last i checked)
Idk how difficult or prohibitive that would be for those with less robust computers though, but since it captures corrections and the entire process, not just the final brush strokes you kept, it seems like the most failproof option.
At least, ai seems incapable of including the corrections, erasing, redrawing, rotating, zooms, etc. That come with screen recs, that the simple timelapses tend to lack
Perhaps a combination of the two would be an almost garuntee that ai would not be able to render identical processes for both the timelapse and screen rec, something to consider so long as it doesn't barr other artists from participating.