OpenNPR. R&D-дневник. Часть 1.

OpenNPR — an open, bilingual (RU/EN) R&D journal on stylized rendering (NPR) and 3D anime pipelines.

3D graphics are not merely a collection of polygons—they represent an endless process of problem-solving. The most compelling stylization techniques today emerge at the intersection of programming and classical drawing. These techniques often remain proprietary within the pipelines of major Asian studios.

The philosophy of open source underpins media. In development, code is customarily shared openly on GitHub, yet in the 3D sphere, valuable knowledge is often locked behind paywalls, hidden by language barriers, or sold through expensive courses. OpenNPR is a quiet, open space for those deeply exploring 3D anime. Here, complex reverse engineering is transformed into accessible instructions.

The media handles the routine tasks: searching foreign resources, transcribing Chinese and Japanese tutorials, and breaking down game models into components. The project saves the most valuable asset of a 3D artist—their time. Knowledge and working files are provided free of charge in two languages.

Global segment—GameDev, Technical Art, and CGI. Niche focus—NPR (Non-Photorealistic Rendering) and 3D-anime pipelines.

Our conceptual benchmark is the English-language YouTube channels that analyze models from gacha games. This is a strong presentation standard, but such videos are lengthy, entertainment-focused, and explain techniques mostly theoretically. In the Russian-language segment, there are no analogs with such deep NPR expertise. Indirect competitors include large CG schools and broad 3D communities—they are expensive and teach fundamentals, classic stylization, or realism.

The Media track is designed for practitioners who have at least a couple of years of experience in 3D. Interfaces and basic modeling are not covered here—the focus is on complex concepts, theory, and implementation.

The first segment is for the 3D practitioner who values their time. While working on a scene, they want to quickly find a ready-made setup for custom normals and apply it immediately, rather than spending ten hours watching Japanese tutorials.

The second is a technical researcher. They are interested in analyzing the character rigging from a new gacha game to understand the pipelines of major studios and optimize their models for real-time performance.

The third is an indie developer looking for understandable NPR shaders for their game’s visuals, relying on ready-made Shader Graph screenshots.

Fourth — the introverted professional. They want to read an author’s blog, quietly download theory and shaders, without intrusive interaction or attempts to sell something.

The project is not confined to a corporate assembly line: it makes room for both technical documentation and personal perspective. The core components include screenshots of Shader Graph (Blender and Unreal Engine) with step-by-step explanations of node logic, video breakdowns of specific techniques, deconstruction of the topology and shading of models from anime and games, as well as free notes, rendering experiments, and source file sharing.

First is reverse engineering of Asian game development. The most challenging part of the research is already complete: we have found references, analyzed model logic, and published a clear algorithm in two languages.

The second aspect is the focus on practice. In media, there are no trite notions or introductory lessons—only advanced shading theory and ready-made technical solutions.

Third—an open-source approach. The media does not monetize every morsel of knowledge: source code, discoveries, and algorithms are freely accessible.

Traditional marketing is unsuitable here. The media platform does not place banners, purchase advertisements, or operate through aggressive sales. In a niche environment, sincere and useful material promotes itself: a specialist looking for an anime shader will click on the headline of an analysis because such information is scarce. The project’s growth relies on the voluntary support of a grateful audience.

Telegram Channel — the primary media platform. The channel format is perfectly suited for detailed breakdowns: GIFs demonstrating shaders are published without quality loss, source files are attached to the posts, and the channel functions as an archive of notes. Every post is released in two languages—Russian and English.

GitHub — the open-source project library. Here you will find the source code (node groups, scenes, shaders) and complete versions of large case studies in the form of articles in two languages. The platform directly extends the open-source philosophy of media: any material can be downloaded, refined, and suggestions can be submitted.

YouTube — a video archive featuring complex analyses that are easier to demonstrate than to describe. The videos are accompanied by Russian and English subtitles.

The plan consists of six modules, each lasting about a month, structured as a sequential course. It begins with understanding how finished NPR characters are constructed. Next, it covers assembling a stylized frame in Compose—a necessary step before the NPR render comes together into an image. Only after that do we delve into specific stylization techniques, ranging from basic shading to the most complex methods. Two posts are released on Telegram weekly: one in-depth analysis and one free note, experiment, or file distribution. The main topic of each module receives a video breakdown on YouTube, and the best materials are polished into full articles on GitHub, complete with source files. Each module concludes with an archival post containing all the files.

Block 1. Deconstruction: How the NPR Character is Built

We start by dissecting finished models from popular gacha games and anime. Topics include face and body topology, auxiliary textures, vertex colors, and material separation. Using these examples, we provide an overview of basic NPR techniques: toon shading, outlines, and hand-drawn shadows. This block functions as a roadmap for the entire course—each technique is covered in detail afterward.

Block 2. Rendering and CompositingAn NPR frame is not generated by a single shader; it is assembled from passes. This block covers custom AOV and render passes for stylization, frame assembly in Nuke and Blender Compositor, color space decomposition and view transform (AgX, Khronos PBR Neutral, ACES via OCIO), and color balancing in the composite.