About AI Usage

This page explains how generative AI may be used as part of the asset creation process for selected products available on this store. Transparency is a core principle, and this information is provided to ensure users clearly understand how assets are developed, refined, and delivered.

The purpose of this document is to describe the role of AI within the production pipeline, clarify its limitations, and explain where human authorship, responsibility, and decision-making remain essential.

General Overview

Assets sold on this store are created using different workflows. Some are developed entirely through traditional manual 3D modeling, while others may involve the assistance of generative AI at specific stages of production.

When used, AI functions strictly as a supportive and exploratory tool, never as a fully autonomous replacement for artistic judgment, technical expertise, or human authorship.

How AI May Be Used

Generative AI may be used during early or intermediate stages of development for exploratory purposes, such as abstract shape generation, structural experimentation, or visual variation testing.

AI-generated outputs are never published as-is. They serve exclusively as raw input and are always followed by human review, selection, correction, and refinement.

The AI-assisted generation process used for these assets is based exclusively on text-based command prompts. No third-party images, photographs, artworks, or visual references are used as input during AI generation.

Outputs are generated, evaluated, discarded, and regenerated repeatedly — often across dozens or hundreds of iterations — until a result reaches an acceptable structural and visual baseline. Only after this stage does the asset proceed to human curation, technical refinement, and final production.

In the rare event that visual reference images are used at any stage of development, such references are exclusively original, author-created materials and are never derived from third-party or copyrighted sources.

Manual Refinement & Production

Every asset undergoes a manual production phase, regardless of whether AI was involved at any point. This may include mesh cleanup, removal of invalid geometry, topology correction, UV creation and adjustment, modifier application, scale normalization, material rebuilding, format conversion, and in-engine testing.

Final assets are evaluated based on usability, technical stability, consistency, and real-world production requirements — not on how the initial exploratory data was generated.

Human Authorship, Curation & Accountability

Regardless of whether AI is involved at any stage, every asset published on this store is the result of human authorship through curation, technical refinement, and final responsibility.

The creator determines which outputs are usable, which are discarded, how assets are structured, optimized, named, packaged, and delivered, and whether they meet professional standards for production use.

Authorship is defined by creative control, refinement decisions, and accountability for the final product, rather than by the initial act of automated generation.

This page is not intended to diminish or devalue traditional manual 3D creation in any form. On the contrary, manual modeling, technical craftsmanship, and artistic decision-making remain fundamental and irreplaceable components of the production process.

When used correctly and responsibly, AI-assisted generation is treated as equally valid and complementary to traditional workflows — never as a substitute for human skill, experience, or creative intent. Whether through manual modeling, assisted generation, or curation, the final outcome always depends on human judgment.

Generative tools cannot replace human creativity, authorship, or responsibility — regardless of whether the process involves traditional modeling, AI-assisted exploration, or post-generation refinement.

What AI Does Not Do

Generative AI does not automatically produce finished, optimized, or production-ready assets. It does not organize files, ensure engine compatibility, validate technical requirements, or make artistic or functional decisions.

These responsibilities remain entirely human-driven and are essential for transforming experimental output into usable and reliable assets.

Textures & Materials

All textures included with these assets are either original works, derived from CC0-licensed resources, or generated with AI assistance followed by human selection, adjustment, and integration. No copyrighted, proprietary, or restricted third-party materials are intentionally used.

AI Labels

Each product clearly indicates its creation method through visual labels:

• AI-Assisted — The asset involved the use of generative AI tools at some stage of the production pipeline.
• No AI Used — The asset was created entirely through traditional manual workflows.

These labels are provided so users can make informed decisions prior to purchase.

Copyright, Curation & Responsible Use

AI-assisted assets are produced through a deliberate, controlled, and human-supervised process. Generation parameters are designed to encourage abstract, non-referential results and to avoid the replication or imitation of identifiable copyrighted works, characters, brands, trademarks, logos, or proprietary designs.

All generated outputs pass through human review. Any result that appears overly specific, derivative, or recognizable is discarded or reworked before release.

It is also important to note that copyright considerations are not exclusive to AI-assisted workflows. Traditional manual 3D modeling is equally subject to copyright law when an asset is created with recognizable similarity to protected works.

The act of manual creation alone does not exempt an artist from responsibility if a result is intentionally or recognizably derived from copyrighted designs, characters, or proprietary assets. In all cases, originality is defined by the final output and its independence from protected source material — regardless of the tools or methods used.

Context on Copyright Risk & Asset Scope

Discussions surrounding copyright risks related to generative AI most commonly involve recognizable intellectual property, such as characters, logos, branded elements, iconic objects, or assets with clear real-world ownership.

This store does not operate within that domain. The assets offered here focus exclusively on non-iconic, organic, and abstract elements, including rocks, stones, trees, plants, monoliths, terrain-like structures, and similar forms that do not have identifiable owners or protected real-world identities.

As a result, the scenarios in which AI-related copyright disputes typically arise are not applicable to the nature of the assets produced and sold here.

In rare and isolated cases, copyright or intellectual property reports may be submitted by third parties. Receiving a report does not automatically indicate an infringement, nor does it result in immediate removal of an asset. All reports are carefully reviewed and evaluated in coordination with the hosting platform, following their established review procedures.

If an infringement is confirmed after proper analysis, the affected asset will be promptly removed from the store and replaced with an alternative asset of equivalent value. All customers who previously acquired the product will be informed about the review outcome and any relevant changes.

It is important to reiterate that all reasonable and possible efforts are taken for each asset. Thorough and careful reviews are performed to minimize any risk of copyright or intellectual property infringement, protecting both the creator and the buyer.

It is also important to note that assets created with AI-assisted workflows are subject to the same copyright and intellectual property considerations as assets produced through traditional methods. AI-assisted creation does not inherently increase the risk of infringement; such risks exist across all creative pipelines and are addressed through the same review, verification, and compliance practices.

Generative Similarity & Organic Forms

In some cases, generative systems may produce outputs that appear broadly similar at an abstract or organic level, particularly with natural forms such as rocks, plants, minerals, or terrain.

Such similarity does not imply copying, derivation, or access to any specific prior work. Without identifiable source material, intentional replication, or a clear chain of derivation, these similarities are considered coincidental and non-actionable under current copyright standards.

Limitations & Good-Faith Disclaimer

While all reasonable and good-faith efforts are made to ensure originality, it is not technically possible to guarantee absolute uniqueness in every conceivable context.

By purchasing or using these assets, you acknowledge that generative tools, like any creative process, involve inherent uncertainty, and that the creator cannot be held liable for speculative or abstract similarities that do not constitute identifiable copyright infringement.

Third-Party Engines & Platform Policies

The use of these assets within third-party engines, tools, or platforms is subject to the terms, policies, and disclosure requirements of those respective platforms.

Compliance with engine-specific or platform-specific AI policies is the responsibility of the end user. The creator is not responsible for changes in third-party rules, enforcement decisions, or platform-specific restrictions.

Legal Status of Generative AI

The use of generative AI tools in creative workflows is not illegal and does not, by itself, constitute copyright infringement. Legal responsibility is determined by the final output and its compliance with applicable laws, not by the mere use of AI during production.

Reporting Concerns

If you believe that any asset available on this store may unintentionally infringe upon your rights or the rights of a third party, please report the issue using the form below.

All reports are reviewed responsibly, respectfully, and in good faith.

Report a Copyright Concern

Quality, Variety & Technical Limitations

One of the main strengths of AI-assisted workflows is the ability to generate a significantly higher level of visual variety when compared to traditional manual modeling. This approach enables a broader exploration of shapes, silhouettes, proportions, and structural variations that would be extremely time-consuming or impractical to achieve through purely hand-crafted processes.

While fully manual modeling allows for absolute control over every individual element, AI-assisted creation prioritizes diversity and experimentation. The trade-off for this expanded variety is the potential presence of minor technical imperfections, such as localized topology irregularities, subtle surface artifacts, texture seams, or small areas of texture bleeding.

It is also important to note that the assets sold in this store are created with a gameplay-first mindset. They are not intended to follow real-world scale accuracy, strict realism standards, or perfectly optimized topology in every aspect. The primary goal is visual readability, performance suitability, and effective use within interactive environments.

Every asset undergoes analysis, refinement, and curation to minimize technical issues as much as possible. However, a perfectly uniform topology — where every edge flow and surface detail is individually constructed and controlled — cannot always be guaranteed in the same way as in fully manual workflows.

These minor imperfections are typically noticeable only under close inspection or non-standard viewing conditions. In real-world usage scenarios, such as gameplay or scene composition at practical camera distances, they are generally imperceptible and do not affect visual quality, usability, or performance.

In practice, this approach provides users with access to a wide range of unique assets designed to function reliably in games and real-time applications, where visual impact and variation are often more valuable than technical perfection.

Retopology & Normal Maps

During retopology, AI-generated meshes are optimized by reducing vertex count and simplifying geometry for real-time use. This process naturally results in a smoother surface, as fine geometric details from the original high-density model are removed.

In many cases, this smooth appearance is beneficial for assets such as plants, rocks, organic forms, and other natural elements, where clean silhouettes and softened surfaces are desirable. However, for assets that rely on sharper surface definition or fine detail, a proper normal map is essential to restore depth and visual complexity.

If normal maps are not supported by the target engine, or are intentionally disabled, the asset will display a fully smoothed appearance. This behavior is expected and should be considered when evaluating the asset for specific use cases.

Metallic & Roughness Maps

Due to the nature of AI-assisted generation and automatic material interpretation, some assets may occasionally include inaccurate or unnecessary metallic and/or roughness maps. This happens because the system does not always perfectly understand which parts of a model should behave as metal, non-metal, or purely decorative surfaces.

In these cases, the affected maps may introduce subtle shading artifacts or darker areas, especially under certain lighting conditions or in specific engines. These artifacts are typically only visible under close inspection and are not noticeable in normal gameplay or scene usage.

All assets go through manual inspection. When such cases are found, they are reviewed and corrected manually whenever feasible. In some situations, the issue is very localized and purely technical, does not affect the overall visual quality, and does not justify discarding an otherwise good and usable asset.

In RPG Developer BAKIN, materials are adjusted by the artist to eliminate almost all visible issues and to provide a clean, production-ready result. When using the assets in other engines or tools, users may simply review, tweak, or disable the Metallic and/or Roughness maps as needed to better fit their own rendering setup.

Mesh Density, Texture Resolution and Performance

In practice, texture resolution has a much bigger impact on performance than the number of vertices in most situations, especially when many objects are visible in the scene at the same time. For this reason, assets are provided with textures in 3072×3072 or 2048×2048 as a quality baseline, but it is highly recommended to resize them to what actually makes sense for your project using the official tool: JMS Texture Resizer.

Good optimization always starts with a simple evaluation of how the object is used: camera distance, how big it appears on screen, and how often it is visible. Small objects such as mugs, buckets, plates, and similar props can easily use 512×512 or even 256×256 textures without any noticeable visual loss in most cases.

It is also very important to check whether an object uses its own unique textures or if it is part of a shared texture set (multiple objects using the same texture files). Shared texture sets are very efficient, but they also mean that any change in texture resolution affects multiple objects at once. For this reason, reducing shared textures too aggressively (especially below 1024×1024) can noticeably reduce visual quality across many assets at the same time.

Textures at 3072×3072 should not be used everywhere. They are just a production standard and can be reduced to 2048×2048 or lower without visible quality loss in the vast majority of projects. Only in specific situations, such as when an object is scaled far beyond its expected size or used in very close-up views, this higher resolution may actually be necessary. As always, the best approach is to follow these guidelines and then validate them with practical in-engine tests. Properly adjusting texture resolution is one of the easiest and most effective ways to improve performance.

Production Process & Workflow

The volume of publicly available assets and the pace at which new content is released did not happen by chance or as a “one-click” result of AI usage. While AI-assisted tools help simplify certain creative steps, the overall production speed is the result of a structured, multi-stage workflow built over time.

This process involves dozens of custom Python scripts and automated tools developed to handle repetitive and technical tasks, such as format conversion, texture processing, mask generation, file renaming, validation, and large-scale folder organization. These systems allow complex asset libraries to remain consistent, organized, and ready for real-time use.

Behind this automation are months of focused experimentation with AI-assisted techniques, combined with years of experience in traditional 3D modeling and asset production. Many workflows were refined through extensive testing, trial and error, and iterative improvements — nothing appeared fully formed or worked perfectly from the start.

AI plays an important supporting role in this pipeline, but it operates alongside planning, scripting, manual review, and continuous adjustment. The final results reflect a balance between automation and hands-on control, ensuring that speed does not come at the expense of consistency, visual quality, or technical responsibility.

AI Usage Disclosure Responsibility

The declaration of AI usage in final projects created with these assets is the sole responsibility of the end user. This store fulfills its role by clearly disclosing the use of AI-assisted techniques during asset creation. Any legal, platform-specific, or policy-related requirements regarding AI disclosure in finished products must be evaluated and handled by the user. The creator cannot be held responsible for issues arising from the omission or misrepresentation of such information in third-party projects.

About Single, Bundle, Collections and Ultimate Editions

To accommodate different budgets and production needs, assets in this store may be available in single packs, bundles, collections, or ultimate editions. This structure exists primarily to give users flexible purchase options and better cost efficiency over time, allowing them to choose between buying only what they need now or securing multiple volumes at a reduced overall cost.

Collections are complete packs designed to provide a sufficiently large and coherent set of assets so users can build their projects without needing to purchase additional volumes. In most cases, a Collection is intended to be a final and self-contained release for that theme. Only in exceptional situations — such as major expansions, new directions, or significant scope increases — may a Collection receive future volumes or extensions.

This store structure was designed based on years of market experience, analysis of purchasing patterns, and long-term observation of what users actually look for when building asset libraries. Each user is free to decide the best way to invest in their projects, while these policies help keep the store sustainable, continuously expanding, and focused on a long-lasting, high-value catalog.

Final Notes

When used, AI exists to improve efficiency and expand creative exploration — not to replace authorship, responsibility, or artistic intent. AI is treated as one tool among many within a broader, human-controlled production pipeline.

It is also important to note that AI-assisted technology is constantly evolving. Our workflow and tools evolve alongside it, and our focus is always to follow these improvements and apply them whenever they bring real, practical benefits. Because of this, all projects are subject to future updates, which may include quality improvements, refinements, and corrections of known limitations as the underlying technology and pipeline continue to mature.

If you are not fully comfortable with the possibility that an asset may have involved AI-assisted techniques at any stage of its creation, it is recommended that you refrain from purchasing such products.

Transparency is provided so that each user can make an informed and confident decision aligned with their own preferences and requirements.