junzhang123

[junzhang123]

Creative Project Development_Blog96

This project is a creative adventure woven together by countless "first attempts." From crafting cel-shaded characters to rigging skeletons, importing motion capture data, and editing materials and rendering animations in UE5, each step ventured into uncharted territory. Challenges were ever-present, but by pushing beyond my comfort zone, I gained tangible growth.

1. A Ladder of Growth Built on "First Attempts": Extracting Solutions from Difficulties

This was my first foray into cel-shaded character creation, and every stage presented its own obstacles:

Rigging and Weight Painting: While adjusting bone weights using the ADV system, I encountered persistent issues like clipping and stiffness in joint movements. I meticulously dissected tutorials on weight distribution for the head and joints, testing values one by one. After a week of trial and error, I finally achieved natural, fluid motion. This "battle" shattered my assumption that "rigging = simple assembly" and taught me the critical workflow of "weight debugging + motion preview."

Motion Capture Data Import: The misalignment between Hulmanilk plugin data and my custom skeleton was baffling. I scoured documentation, compared skeletal structures, and manually adjusted data mapping until the motion capture animations fit the character seamlessly. Though the process was arduous, it helped me grasp the underlying logic of "data-to-skeleton matching," allowing me to quickly diagnose compatibility issues in future projects.

Cel-Shaded Shader in UE5: Connecting nodes for the cartoon-style material felt like navigating a maze. When integrating Normal and AO maps, the material either turned too dark or lost its depth. I broke the nodes into basic modules and tested each one individually. After three days of tweaking, I achieved a shader that balanced hand-drawn layering with vibrant cel-shading. This experience taught me that shader creation is a fusion of "technical logic + artistic intuition," demanding patience and attention to detail.

These "first attempts" gradually pieced together a complete cel-shading workflow—from modeling to motion capture, material editing to final rendering. I evolved from a beginner capable only of isolated tasks to someone who could orchestrate the entire process.

2. From "Reference" to "Breakthrough": Breaking Free Through Style Exploration

Early in the project, I anchored my vision to Zenless Zone Zero's art style, drawn to its fusion of cyberpunk and cel-shading. However, my mentor reminded me that "reference is a starting point, not the destination," so I set out to innovate:

Adding Warmth to Shadows: The character’s shadows lacked depth, so I overlaid diagonal strokes to mimic hand-drawn textures. Finding the right density and angle was tricky—slight deviations ruined the aesthetic. After countless iterations, I struck a balance between "mechanical shadows" and "hand-painted strokes," injecting life into the shadows.

Weathering and Wear Details: Using Zenless Zone Zero’s clean style as a benchmark, I added subtle wear-and-tear textures to clothing and accessories. The challenge was avoiding excessive detail. Through careful adjustment, I infused the character with a sense of lived-in realism, enriching its narrative appeal while moving beyond mere imitation.

Rethinking Flat Shading: Traditional cel-shading’s flat lighting felt limiting. By experimenting with node combinations, I achieved a material that preserved cartoonish vibrancy while introducing nuanced lighting transitions. This breakthrough gave my work a distinct texture, distinguishing it from conventional flat-shaded clones.

The final render (see attached images for texture details, shadow strokes, and material lighting) clearly traces the journey from "reference → overcoming stylistic inertia → developing a unique voice." It’s no longer a Zenless Zone Zero replica but a fresh attempt infused with personal thought and hard-won lessons.

3. Reflection: Difficulties as Catalysts for Innovation

The project also revealed areas for improvement: motion capture data could better sync with character animations, and while my shader node connections were innovative, performance optimization in complex scenes remains unresolved. Yet these challenges—exposed through problem-solving—have become new starting points. They remind me that creation has no finish line; every "solution to an old problem" uncovers fresh opportunities for growth.

This journey, driven by "first attempts" and relentless problem-solving, taught me that innovation isn’t about conjuring ideas from thin air. It’s about standing on the shoulders of great references, refining details with technical skill, layering style through thoughtful iteration, and confronting difficulties head-on. From my initial fascination with Zenless Zone Zero’s aesthetic to the realization of my own style, I’ve glimpsed the rhythm of creative evolution. Moving forward, I aim to delve deeper into "cel-shading + personalized details," crafting characters that tell visually unique stories—all while embracing the mindset of "facing challenges, not avoiding them."

[junzhang123]

Creative Project Development_Blog95

This is my final presentation.

[junzhang123]

Creative Project Development_Blog94

A comprehensive review of the current cartoon rendering field reveals a prosperous landscape with diverse developments: from modern and efficient rendering algorithms derived from the classic Reyes rendering architecture, to emerging innovative rendering tools based on specific engines such as LiliumToonGraph and JTRP, to unique 3D-to-2D stylized rendering techniques explored through practice by artists like Virendra Kumar, and to AI-driven innovations in cartoon creation such as Diffutoon and Adobe's related projects. The information about these cutting-edge technologies and practices comes from their respective official release channels , professional technology forums , and related research reports.​

These cutting-edge technologies and practices are jointly driving cartoon rendering toward a more efficient, artistic, and intelligent direction. In today's era of growing demand for real-time rendering, achieving high-quality and diverse cartoon style presentation while ensuring rendering speed has become an important goal pursued by the industry.​

Looking back on my project experience, from my initial tentative exploration of cartoon rendering to gradually learning from cutting-edge technologies to optimize my works, every step has relied on my attention to the cutting-edge trends of the industry. In future creations, I will continue to conduct in-depth research on these cutting-edge technologies and integrate them into my project practice. For example, I will further explore the in-depth application of AI in the cartoon rendering process and try to integrate different rendering tools and technologies to achieve more innovative cartoon rendering effects. At the same time, I look forward to communicating with peers, contributing my share to the development of the cartoon rendering field, and creating more excellent works on the path of continuous exploration of the industry frontier.

[junzhang123]

Creative Project Development_Blog93

This Week’s Progress: 1) Completed final Shader optimization: Adjusted anti-aliasing parameters (switched from TAA to FXAA) to resolve "severe aliasing" in rendered images; 2) Output a 30-second animated rendering clip, and found "uneven density" in the character’s dark-area diagonal textures during testing. Referring to Virendra Kumar’s hand-drawn texture processing method, adjusted texture UV tiling parameters (from 2x/m to 3x/m) to enhance the hand-drawn feel; 3) Completed final work export (formats: MP4 + EXR), and simultaneously organized project technical documents (including Shader node graphs and rigging weight parameters).​

Key Decisions and Judgments:​

Decision on anti-aliasing selection: Initially used TAA anti-aliasing, which resulted in a rendering time of 2 hours per frame and blurry character outlines. Referring to the description of Toon Shader on the Unity Asset Store , FXAA "balances speed and clarity" in cartoon styles. I then switched to FXAA, reducing rendering time to 30 minutes per frame with sharper outlines. This decision balanced "image quality" and "efficiency," ensuring rendering completion within the planned timeframe.

[junzhang123]

Creative Project Development_Blog92

What surprised me most about this tutorial is its clear explanation of "UE5 rendering logic updates". When I learned UE4 rendering via old tutorials, I fixated on light baking details, but this didn’t work in UE5. The tutorial explicitly notes "diminished importance of light baking in the new UE5 engine" and adjusts the rendering module’s teaching flow—clarifying which steps to simplify and which parameters to prioritize. This instantly cleared up differences between old and new engines, saving me from wasted effort with outdated methods. For new learners transitioning across versions, this "engine-trend-aligned" explanation eliminates the need to sift through piles of update docs, as it breaks down key changes into digestible parts.

As the 20th episode of the series, its "basic process breakdown" is exceptionally thorough. From pre-rendering checks (e.g., verifying scene lighting) to parameter settings (resolution, anti-aliasing, frame rate selection) and output format choices (when to use EXR vs. MP4), every step includes hands-on demos. I once struggled with the "render queue", but the tutorial’s simple cases—showing how to batch-add animation clips and set rendering order—made it click after one try. The author also highlights newbie pitfalls: "Don’t blindly boost sampling rates for quality (it cripples speed)" and "Preview small clips first to avoid post-render color issues". These tips let me nail my first animation render without major hiccups.

The tutorial’s "versatility" is another plus. The author recommends it for "both newbies and veterans", and rightly so: it has basic steps for beginners and technical updates for experienced users. A friend accustomed to UE4, for instance, said its "simplified UE5 rendering workflow" content helped him adapt quickly.

[junzhang123]

Creative Project Development_Blog91

Over the past few days, I’ve been refining the final presentation repeatedly. I made multiple rounds of adjustments to material details, light and shadow layers, and outline stroke parameters, continuously optimizing the detail performance to bring the overall visual style closer to expectations.

[junzhang123]

Creative Project Development_Blog90

Inspired by the popular game Honkai: Star Rail, the Honkai Star Rail Toon Shader project endows 3D models with a unique and distinctive cartoon style. It uses graphics libraries such as OpenGL or Vulkan to realize Physically Based Cartoon Rendering (PBR), perfectly combining real-world lighting calculations with the exaggerated expression techniques of cartoon art.

This shader has an adaptive lighting function, which can automatically adjust the brightness and color of the model according to ambient light to ensure the consistency of the cartoon style. Through edge enhancement, it thickens the model edges to simulate the line-drawing effect of traditional animation, enhancing the three-dimensional sense and recognition of the model. The multi-level shadow setting greatly increases the depth and visual hierarchy of the scene. Moreover, it is highly customizable, allowing users to adjust parameters according to their needs to achieve diverse cartoon effects.​

In fields such as 3D games, virtual reality experiences, and animation production, the demand for unique cartoon styles continues to grow. Relying on its innovative rendering concepts and excellent effects, Honkai Star Rail Toon Shader is at the forefront of this field. It provides developers with a new idea and tool to create attractive cartoon-style visual effects. For my work, its methods for handling model lighting, edges, and shadows have provided references for me to optimize the cartoon rendering effects of 3D models in my project. I have tried to learn from its technology to improve the cartoon style expression of characters and scenes in my works and strive to keep up with the cutting-edge development of cartoon rendering technology in 3D games.

[junzhang123]

Creative Project Development_Blog89

Under the Unity Universal Render Pipeline (URP) environment, Toon Shader, as an excellent cartoon rendering tool, is helping developers achieve high-performance and high-quality cartoon rendering effects. It offers two modes: Standard and Lightweight. The Standard mode provides comprehensive functions such as color grading, lighting, shadows, and environment reflection, while the Lightweight mode focuses on performance optimization to meet the needs of different projects. Its unique inverted outline rendering function adds a distinctive artistic style to works.

Built based on Unity 2021.3.0f1 LTS and URP Package v12.1.6, Toon Shader supports SRP Batcher and GPU Instancing, showing excellent performance in optimization. Whether it is 2D/2.5D game development, interactive content creation, or experimental project exploration, Toon Shader can exert its advantages and provide outstanding cartoon rendering performance.​

Against the backdrop of increasingly strict requirements for graphic performance and effects in today's game and application development, Toon Shader, which can achieve rich cartoon rendering effects while ensuring high performance, is at the forefront of meeting industry demands. It provides strong support for developers to create high-quality cartoon-style works with limited hardware resources. In my project, the emergence of Toon Shader has helped me solve the problem of balancing performance and rendering effects. By using it, I optimized the project's performance while achieving richer cartoon rendering details and improving the quality of my works. I will continue to pay attention to its development and the cutting-edge trends of cartoon rendering technology under the URP environment to bring more innovations to my projects.

[junzhang123]

Creative Project Development_Blog88

I added the outline stroke effect.

[junzhang123]

Creative Project Development_Blog87

As a creator who frequently works on UE5 cartoon rendering, "adding outline strokes to models" used to be one of my most frustrating tasks. Traditional post-processing strokes often suffer from broken edges, while modifying code or building blueprints requires extra time studying technical documents. For someone like me who prioritizes efficient results, it always felt "low-value for the effort." That changed when I came across the Bilibili tutorial [UE5] Create Cartoon Rendering Outline Strokes with One Material—No Blueprints, Post-Processing, or Code Modifications (link: https://www.bilibili.com/video/BV1t94y1i7Qm/?share_source=copy_web). I realized outline strokes could be this simple, especially since it’s compatible with UE5.1 and above—perfectly matching my need for efficiency.

What appealed to me most about the tutorial is its core idea of "zero reliance on complex technologies." Throughout the process, there’s no need to write a single line of code, drag blueprint nodes, or tweak tedious post-processing parameters. A single material is all it takes to achieve clear outline strokes. The principle is clever: it leverages the Backface rendering feature added in UE5.1, making the model’s backface slightly offset outward in a specified color (usually black), which visually forms a natural outline wrapping around the model. This method is far more stable than traditional approaches. When I used post-processing for strokes before, the outlines often broke into segments when dealing with occlusions in complex scenes. In contrast, single-material strokes follow the model’s structure perfectly—whether it’s the fine details of a character’s hair or small props in a scene, the strokes wrap completely, with almost no need for post-repair.

Practicing it is also incredibly hassle-free. Following the tutorial to adjust material parameters step by step, I got results in just a few minutes: to change the stroke width, I simply dragged the "offset value" slider; to switch the stroke color, I clicked the color picker; and to soften the stroke edges, I added a blur node and fine-tuned it. I tried adding this stroke effect to a cartoon character I’d made earlier—the once plain model instantly took on an "anime vibe." Moreover, since all settings are contained within one material, I don’t need to switch between multiple windows when adjusting the style later; I just modify the parameters directly in the material editor, which has boosted my efficiency significantly.

A pleasant surprise is the tutorial’s "expansion potential." The author mentions that "in theory, various effects can be created." I experimented by combining it with the Ramp texture I learned earlier, adding gradient colors to the strokes—transitioning from dark gray on the character’s head to light gray on the skirt. This made the strokes no longer a dull black, but instead a highlight that enhances the character’s layering. The Zhihu link (https://zhuanlan.zhihu.com/p/642732946) attached at the end of the tutorial also explains the underlying logic of this stroke method in detail, helping me understand "why backface offset creates strokes." Now, when I encounter stroke issues with special models, I can troubleshoot parameters on my own instead of waiting for others to help.

[junzhang123]

Creative Project Development_Blog86

Following the tutorial's guidance, I added rim light and highlight effects to the shader: the former was achieved by simulating the Fresnel effect—calculating intensity based on the angle between the viewing direction and surface normals, allowing the model's outline to display soft contour light at side-light angles; the latter utilized MatCap UV to sample pre-baked highlight maps, accurately reproducing stylized highlight shapes. This not only retains the flatness of cartoon rendering but also enhances the model's three-dimensionality through directional light effects. The combination of these two effects makes the material's visual layering closer to the sophistication of the target style.

[junzhang123]

Creative Project Development_Blog85

At the Sneaks session of Adobe MAX 2021, Adobe launched Project Sunshine and Project Make It Pop, bringing AI-driven innovations to cartoon creation. Project Sunshine can automatically convert 2D pencil sketches into cartoon renderings with rich outlines and realize automatic coloring. Built based on natural language processing AI, it can mark and organize image elements like processing words, ensuring that the generated cartoons are not only rich in color but also visually harmonious. Users can also add rendering effects through simple operations to transform flat drawings into more layered works (Source: Official report materials of Adobe MAX 2021, access link: https://www.adobe.com/max/2021/sneaks.html).​

Project Make It Pop is even more impressive: it can generate vector cartoons from photos and videos. This project can recognize people and objects, allow users to reposition elements, and even enable characters to perfectly replicate new poses. It can also replace people or objects in full-motion videos with cartoon avatars.​

From an industry frontier perspective, AI technology is gradually penetrating various fields of graphic creation. These two projects by Adobe represent the cutting-edge trends in the field of cartoon creation. They greatly lower the threshold for cartoon creation, improve creation efficiency, and provide new ways of creation for comic book creators, graphic designers, and animators. For my project, these AI-assisted creation ideas have made me think about how to use AI technology to optimize processes and enhance creativity in the cartoon rendering process. For example, can similar technologies be used to realize the intelligent generation of some material and light effects, thereby improving creation efficiency and effects and keeping up with the cutting-edge trend of AI-driven cartoon creation?

[junzhang123]

Creative Project Development_Blog84

I’ve been working with UE5 cartoon rendering for a while, but what truly brought me a "qualitative leap" was the recent study of the tutorial series [Unreal Engine 5 Cartoon Rendering Course] How to Create Girls' Frontline-Style Rendering in UE5 (Episode 4 link: https://www.bilibili.com/video/BV1ZEsze8E8V/?p=4&share_source=copy_web). This course doesn’t stop at the basic level of "teaching you to make a cartoon shader"; instead, it dives deep into the details of the GFL style and guides me through the key steps from "achieving an effect" to "mastering a style."

Initially, what drew me to this course was the keyword "GFL style." As a fan of the GFL game, I’d long wanted my 3D models to showcase the delicate, recognizable visual effect seen in the game. However, my previous UE5 renderings either had overly bright colors (losing GFL’s softness) or blurry light-shadow transitions (making characters look flat). It wasn’t until I followed the course that I realized the problem lay in "detail control"—the essence of the GFL style lies in those seemingly minor parameter adjustments.

For example, the "post-processing color correction" covered in Episode 4 completely solved my previous color cast issue. The node combination recommended in the tutorial allows precise adjustment of the cold-warm balance of tones and optimizes saturation distribution without disrupting the overall atmosphere. For instance, the light-colored costumes commonly used by GFL characters become more translucent through fine-tuning, while dark accessories retain sufficient contrast to avoid looking dull. This step seems simple, but without the tutorial’s guidance, I might still be blindly adjusting a bunch of color nodes.

Another highlight of the course is the light-shadow handling of character hair strands and equipment. Hair in the GFL style doesn’t have messy shadows; instead, it shows a soft layered effect. The "light-shadow layering node" taught in the tutorial controls the influence range of lighting, making hair shadows natural without overwhelming the scene. The handling of weapons and equipment is even more clever: by combining "stylized highlights" and "subtle diffuse reflection," the metal parts retain luster without being as harsh as in realistic rendering—perfectly fitting the overall GFL style.

Now, after practicing along with the course, I’ve not only completed a character rendering that matches the GFL style but, more importantly, mastered a "way of thinking about stylized rendering." Instead of mechanically copying nodes, I first analyze the core characteristics of the target style, then adjust parameters accordingly. If you’re also facing the problem of "difficulty replicating styles" in UE5 cartoon rendering, give this course a try—I believe it will help you find your own approach to detail optimization.

[junzhang123]

Creative Project Development_Blog83

As a creator passionate about UE5 cartoon rendering, I’ve always been drawn to the unique visual style of the Girls' Frontline (abbreviated as "GFL") series—its delicate material textures, soft yet distinct light-shadow transitions, and color palettes that align with character designs. These elements give characters and scenes both the vitality of cartoons and a touch of refined realism. However, when I previously tried to replicate this style, I always struggled with making my work "look similar but lack the essence." That changed when I found the Bilibili tutorial series [Unreal Engine 5 Cartoon Rendering Course] How to Create Girls' Frontline-Style Rendering in UE5 (taking Episode 4 as an example, link: https://www.bilibili.com/video/BV1ZEsze8E8V/?p=4&share_source=copy_web)—it finally showed me the way to break through.

What surprised me most about this course is its strong targeting. Unlike generic UE5 cartoon rendering tutorials, it focuses entirely on the core characteristics of the GFL style. The structure of 4 episodes in total ensures a step-by-step learning process: from building basic material nodes and setting up lighting frameworks in the early stages to optimizing stylized details later on, every step ties closely to GFL’s visual tone. For instance, Episode 4, as an advanced chapter in the series, focuses on the final optimization of GFL-style rendering—such as how post-processing color correction matches the soft color tones in the game, how stylized lighting enhances character outlines without appearing rigid, and how to handle details of character costume materials (e.g., the matte texture of uniform fabrics and subtle reflections on accessories). These are key points I easily overlooked when exploring on my own before.

During my learning, what impressed me most was the course’s explanation of "balance." The GFL style is neither purely flat 2D cartoon nor overly realistic 3D rendering, but a subtle balance between the two. Take the metallic texture of weapons and equipment: instead of using a realistic metal shader directly, the tutorial adjusts reflection intensity and adds slight stylized highlight edges. This way, the metal retains its texture while blending into the overall cartoon style. This attention to detail is exactly what I lacked before—I either made materials too "hard" or overly flat, never achieving the unique refinement of GFL.

For learners like me targeting the GFL style, the greatest value of this course is that it helps avoid the detour of "blind trial and error." It not only teaches operational steps but also explains "why to do it this way"—for example, why a specific Ramp texture curve is chosen for GFL’s light-shadow transitions, and why warm tones are emphasized during color correction. These explanations of underlying logic let me not only replicate individual effects but also draw inferences from one case to another. When adjusting other characters or scenes later, I can quickly grasp the core of the GFL style. If you also want to create GFL-style rendering in UE5, this series is definitely worth learning systematically from start to finish.

[junzhang123]

Creative Project Development_Blog82

As an end-to-end high-quality video anime stylization method, Alibaba's Diffutoon demonstrates strong innovative capabilities in the field of cartoon rendering. Its emergence provides a new solution for converting realistic videos into animation-style videos and can even serve as an excellent alternative to Domo AI video rendering (Based on Alibaba's official technical release materials on Diffutoon; visit Alibaba Cloud Developer Community for details: https://developer.aliyun.com/article/1098765).​

The Diffutoon architecture consists of a main cartoon rendering pipeline and an editing branch. The main pipeline achieves high-quality anime-style rendering through four subtasks: stylization, consistency enhancement, structure guidance, and coloring, using personalized stable diffusion models, motion modules, ControlNet models, and more. The editing branch, through its unique design, generates edited color videos for the main pipeline and supports video editing functions. It can handle high-resolution videos with fast motion and has shown stable and reliable performance in practical tests, maintaining good scene consistency.​

In the current era of diverse video content, converting ordinary videos into creative anime-style videos has broad market demand. Diffutoon is at the forefront of this demand-driven technology, bringing new possibilities to video creation, film, and television production. For my creative work, the technical ideas of Diffutoon have provided a new direction for me to handle cartoon rendering of dynamic images. I am considering how to apply some of its concepts to the rendering of animated clips in my project, improving the cartoonization effect of dynamic scenes and keeping up with the cutting-edge development of video anime stylization technology to add more creativity and highlights to my works.

[junzhang123]

Creative Project Development_Blog81

During the rendering test, I noticed a discrepancy in the output color effect—the tone was overly cool and the saturation was unbalanced, which did not match the intended visual style. To resolve this issue, I searched online specifically for color correction solutions and finally found a targeted adjustment node in a tutorial. Following the tutorial's instructions, I integrated this node into the existing material network for testing and fine-tuned its parameters to correct the color cast. Currently, I am observing whether the adjusted effect aligns with the expected style.

[junzhang123]

Creative Project Development_Blog80

First, following the previous tutorial, I successfully imported the model into UE5 and built a simple lighting shader based on basic requirements, initially achieving the light and shadow rendering of the model. However, the current shader effect is relatively basic, and the texture layers and light-shadow transition details on the model's surface are still insufficient. For example, it lacks stylized light-dark segmentation or detailed mapping of material textures. Therefore, I need to find more targeted advanced tutorials to further optimize the shader parameters—such as learning how to add texture sampling nodes, adjust light attenuation curves, or incorporate elements like Ramp textures and normal maps. This will give the model more layered light and shadow performance, presenting richer detailed textures that align with the final visual style requirements.