My topic about Krita Smart Select was removed with the reason “Promotes or Includes AI-Generated content”. I would like to ask for a review or at least a clarification of the rule being applied here.
Krita Smart Select is not an image generator and does not create artwork, replace the artist’s work, or produce AI-generated images. It is a local selection/matting tool: the user draws a lasso around an object, then the plugin creates a soft alpha selection inside that user-defined region. The result is a Krita selection mask, not generated visual content.
I understand that the forum has a policy against promoting AI tools, and I respect that the moderation team may decide what is appropriate for this space. What I am asking about is consistency and classification.
There is currently a public topic in Resources > Plugins and Scripts for “Vision Tools (Selection, Background removal, …)”:
That topic describes tools using Computer Vision models for semantic object selection, background removal, high precision masks, and GPU support. From a user-facing perspective, Krita Smart Select belongs to the same broad category: local computer-vision-assisted selection/masking for Krita, not image generation.
So I would appreciate clarification on one of these points:
Are local computer-vision selection/matting plugins allowed when they do not generate images?
If not, why is the Vision Tools topic still available in the Plugins and Scripts category?
If my post wording made it sound like a generative AI plugin, could I edit it to make the scope clearer?
If the post cannot be restored, could the moderation team clarify the rule publicly so plugin authors know whether non-generative CV tools are allowed or not?
I am happy to adjust the post to avoid any misleading wording, remove promotional language, and clearly state that the plugin does not generate images and only creates user-controlled selection masks.
I am not asking the forum to endorse my plugin. I am asking for the rule to be applied consistently. If “AI” means any machine-learning-assisted local tool, then existing public plugin topics in the same category appear to violate the same rule. If the distinction is different, plugin authors need to know what that distinction is.
Yes, the plugin uses a local machine-learning model, BiRefNet HR Matting, to produce a soft alpha selection mask from a user-drawn lasso region.
It does not generate images, artwork, textures, styles, prompts, or any visual content. It only refines a user-controlled selection area into a Krita selection mask, similar in purpose to foreground/background selection or background removal tools.
That is exactly why I’m asking for clarification. If the forum policy now prohibits any plugin that uses any ML/AI dependency, including non-generative local computer-vision tools, then I understand that my post is not allowed.
But in that case I would like to understand why this existing topic is still publicly available in Resources > Plugins and Scripts:
That plugin also describes computer-vision/model-based selection and background removal functionality for Krita.
So my question is not whether my plugin uses ML. It does. My question is whether non-generative local computer-vision selection tools are allowed on Krita Artists, and if not, whether this rule is being applied consistently.
Thank you for the clarification. Part of our objection to AI tools is the LLM using others’ work without permission. The Krita Fast Lineart tool was trained on artwork submitted by the artists who decided to contribute images so that’s why it’s allowed. The other tool you brought up… I think we have to look at that one again. This is all evolving and changing (and people like me still have a lot to learn) so give us a little time to respond.
Just to make sure I understand the boundary correctly: Krita Smart Select does not use an LLM, does not generate images, and does not create artwork from prompts. It uses a local computer-vision matting model to turn a user-drawn lasso region into a Krita selection mask.
The model currently used is BiRefNet HR Matting. I can provide links to the model page, license, and technical details if that helps the moderation team evaluate it.
For me, the main question is what the forum policy is meant to prohibit:
generative AI / LLM-based tools specifically;
models trained on data without clear contributor permission;
or any plugin with any ML/AI dependency, including non-generative local computer-vision tools.
I’m fine waiting for the moderation team to discuss it. I just want the boundary to be clear, because it affects whether plugins like Smart Select, Vision Tools, and similar selection/background-removal tools are allowed to be shared here at all.
Code (plugins and scripts) that is AI generated will also not be allowed anymore.
This is the case for the plugin, meaning it would not be allowed on those grounds already. Not really a need to figure out what it does if it fails this criterion already.
I would like to separate two different issues here.
The first issue is whether the plugin is allowed because it uses a local ML model for selection/matting. That is the topic I was already asking about.
The second issue is the new claim that the plugin code itself is AI-generated. I do not think that claim should be made without evidence.
If this conclusion is based on the presence of AGENTS.md in the repository, I would like to clarify that AGENTS.md is not a “Generated-By” or “Co-Authored-By” marker. It is a contributor/agent instruction file, similar in purpose to project documentation for development conventions. Its presence does not prove that the code was generated by AI.
This distinction matters because in the linked policy discussion, moderators themselves noted the risk of false positives and a hostile “prove you are innocent” environment. The thread also mentions concrete signals such as Assisted-By / Generated-By / Co-Authored-By trailers, explicit author statements, or a README that clearly states GenAI usage. I do not think AGENTS.md alone meets that kind of standard.
So I would appreciate clarification:
What specific evidence is being used to conclude that this plugin code is AI-generated?
Does the forum consider the mere presence of AGENTS.md sufficient evidence?
If partial AI assistance is what matters, where is the boundary? Does any use of Copilot-like autocomplete, an AI code review, generated snippets later reviewed and rewritten by the author, or agent-assisted project documentation all count the same way?
What standard of proof is moderation applying before labeling a plugin as AI-generated?
Once again, I am not asking the forum to host something against its rules. I am asking that a separate accusation about code authorship not be treated as established fact without a clear criterion.
If the moderators want my reasoning, they can send me a private message, but I’m confident in my statement. The rest of this post is weaseling and misdirection, it makes up a bogus claim and proceeds to pick it apart.
I’ll mute this thread, this behavior is not worth my time.
Aaaand another baseless claim that not only was the code generated using AI, but so were my responses.
So far you haven’t clarified any of your claims, those being:
Using a computer vision model constitutes as a violation of the forum rules regarding LLMs and generative AI, despite the model not being an LLM and not generating anything besides the selection mask.
The plugin code being AI-generated.
My replies being AI-generated.
I have nothing against moderators banning forum threads based on a whim, but I’d suggest you be a little more upfront and honest about it. Because given the nature of your response I fail to understand the purpose of rules and community guidelines on this forum.
The topic that sooz linked makes the rules pretty clear I think. It also gives an example of what can stay.
The reason the other plugin was still on the forum was simply because I missed it when deleting topics after we extended the rules. It is gone now too.
You are still allowed to do any AI plugin you want. It’s just not allowed to post them here. At least not as long as the Krita foundation doesn’t tell me to revert the ban.