Nausicaa's Secret Garden: My Journey into Blender 3D
- sn1zz3d
- Mar 30, 2021
- 9 min read
Updated: Dec 6, 2021
This is an in-depth analysis detailing my journey in recreating Nausicaa's Secret Garden in Blender 3D. It involves what I learned, the challenges, alternatives, solutions and finally my opinions on the matter. If you're after a shorter overview, I have created a blog post which you can find here.
This piece was my first dedicated project in Blender. I had dabbled in Blender occasionally, but I hadn't considered utilizing Blender's features fully to appreciate it until now.
Image: Obligatory Blender Guru Donut/Mug render I created previously.
Having said that, I will preface by saying I am no expert in the matter of Blender. In fact, most of what I've learned is from Blender's documentation and other artists. I have therefore compiled a list of the amazing artists which have helped me learn the software and allow me to continue my growth in it. If you're serious about learning Blender, I recommend checking them out!
Blender Reference Manual
Blender Tutorials - Blender Guru
Texturing Tutorial - Southern Shotty
All 80+ Blender material nodes explained in under 30 minutes - Daniel Krafft
Mantaflow Tutorials (Fluid + Flame) - CrossMind
Post Process Tutorial - Gleb Alexandrov https://www.creativeshrimp.com/post-processing-tutorial-in-12-simple-steps.html
Improve Workflow Efficiency - Blender Secrets
---------------------------------------------------------------------------------------------------------------------------
#0 - Starting Point
I wanted to recreate this iconic scene from Hayao Miyazaki ultimately to challenge my ability with Blender and also to find out if it was possible to enhance the legendary Studio Ghibli aesthetic into the 3D realm.
Looking at the scene in motion there are notable elements that I wanted to capture: the motion of the water, flame flickering from the lantern and warm, glowing look and feel. Later down the line, I ended up giving my personal flair to the scene.
#1 - Inspiration & Reference Gathering
Inspiration largely came to me as I went along, but the notable pieces of media I found were Lupin III: The First (2019), Genshin Impact and Dishonored 2.
Fun fact: Hayao Miyazaki directed Lupin III: Castle of Cagliostro
The richly detailed and painterly look of Dishonored 2 combined with the densely saturated color depths of Genshin Impact felt like good directions to take a Studio Ghibli piece into 3D.
I am a huge advocate for developing a plan, whether it be roughly outlined or detailed (fail to plan, plan to fail), so for the first week I gathered references, drew out what kind of assets I'd create and the processes I needed to get there. I used PureRef which is free program allowing users to consolidate images into a giant ever-expanding board!
The screenshot of Nausicaa's secret garden is a feast for the eyes with seemingly innumerable objects to create. In order to alleviate what appeared to be a colossal undertaking, I grouped up assets into groups which share some similarity.
This is one of the most important steps because It meant creating approx. 14 unique assets, altering their geometry, shading and positioning in order to give the impression of diversity.
#2 - Setting up the Scene (Blockout)
The first step was setting up the scene. This involved position the crucial reference image which I would assist in the modeling process. Unfortunately, using this reference image introduced a problem that'd persist throughout the project length...
The angle and perspective of the scene from Nausicaa, turned into an elongated environment when viewed from the side or top. I had two choices, I could disregard the reference image and create properly scaled environment or adhere to the reference image completely. I chose the latter ultimately to reinforce a stronger comparison between the reference and the final render.
Image: Notice the distance between flora when viewed from the side.
#3 - Modeling & UV Unwrapping
For modeling, It's good practice to start with low-poly primitives and build detail upon them over time. Personally, I like to build the entire scene with basic shapes and then enhance them altogether. This helps to get an idea for the direction of the scene when you look back on it.
I applied a handful of different modeling techniques from boolean, subdivision, retopology, curve modeling and box modeling. If you ever think what modeling technique should you use, just ask yourself if it would be easier or more time-effective to model this way or not.
Image: Assets with basic shapes were left out when I determined using particle systems and sculpting were more suitable.
Once I was happy with the level of detail in the geometry, I proceeded into unwrapping the UVs. Since I didn't have to worry about viewing the scene in 360 degrees, I was able to get away with seams in my texture work. It is always recommended to think about how your assets will be viewed to figure out if you can reduce project length.
Image: Seams visible from the backs of most assets.
#4 - Duplicate Linked & Particle Systems
One of Blender's awesome features are linked duplicates. This essentially means you can make edits to one asset so long as they are linked. This has saved me heaps of time when I found mistakes in my mesh or if I wanted to enhance a part of the geometry. To link duplicates, you can select a mesh and press alt+D. There are many other ways to link assets together (refer to Blender's Reference Manual for more info).
Image: edits being made to one asset are linked to another (linked duplicates)
The spores and leaf assets were created using particle systems. I modeled out the leaves underneath the scene just to make it easier for me to find them. The spores themselves were created using hair properties particle system. In my opinion, If you can achieve the same look using textures/materials or simple geometry, I'd do that as liberal use of particle systems can increase render times.
Image: the base leaves which will be used as the main object for particle systems.
#5 - Texture Painting & Material/Shading
To be honest, this was the most anxious part of the design pipeline. I was hesitant in hand-painting the textures for the environment as I was unsure whether I could properly grasp the Studio Ghibli aesthetic. Eventually, I coerced my insecure self to paint one texture at a time.
Surprisingly, what seemed initially daunting, turned out to be quite enjoyable! Some of the textures were created in Blender but eventually I made the rest using Clip Studio Paint as I was more comfortable in that software and found that there was more control and tools to use. Not being able to paint in layers is among them**. Texture painting is perhaps one of the aspects Blender could improve upon and I'm hoping that a future update can do just that.
**From the time of finishing this project, I found out there is an add-on which allows users to paint in layers. I cannot comment on its viability as I haven't used the add-on but you can find a link to it here.
Image: seamless textures made in Clip Studio Paint. Click here for an awesome seamless texture tutorial by Rajawat.
Analogous colors were used to minimize unusual color variations during the post-process stage (compositing).
For some assets, I felt I didn't need to hand-paint unique textures and instead utilize Blender's material editor to create finished looks. After learning about what I can do with material nodes, I was addicted to them! Keep in mind that material nodes contribute to render times so it is recommended to 'bake' nodes into a texture only if you no longer need the capability of altering them.
Image: Using a mix shader node with animated UV keys in order to create the impression of flowing water direction.
Image: Texture mask used to denote the direction of the water (white=vertical, black=horizontal).
#6 - Sculpting
The complexity of the two tree-like assets prompted me to sculpt them. I tried out 2 methods of baking down the details into texture normals: Traditional baking using low-poly and high-poly and Blender's multi-resolution modifier. Personally I liked using multi-resolution modifier for baking as it also enabled me to increase or decrease level of detail sculpting and the asset geometry.
One important lesson I learned when sculpting is that Dyntopo should generally be used for sculpting out the shapes of an object or for when you're concepting. You'll notice a significant slow-down when using Dyntopo over time. Honestly, you'd be better off using remesher or multi-resolution modifier in that regard.
Image: High poly bake to low poly.
Image: Multi-resolution modifier bake.
#7 - Fluid & Flame Simulations (Mantaflow)
I initially wanted to create the water streams using Mantaflow simulations but defaulted to animating normal maps. I was not happy with how it looked and most likely it is because of my lack of experience in simulations (see below images for example).
There's not a lot I can say about this section as I found it quite challenging. I can mention a couple of things that can help you from pulling any hairs out of frustration or lack of comprehending.
- If you are using Replay to cache simulations, you must change the resolution of the domain and reset the timeline's frame position to the beginning in order to see the changes to a domain and flow asset.
- Certain settings such as noise (not to be confused with noise artefacts in render) are directly proportionate to the settings in flow assets. If you can no longer see your simulations, try increasing or decreasing some values in the flow asset to determine if that's the case.
- I highly recommend checking out CrossMind's Mantaflow Tutorials on Youtube if you want to learn how to use this powerful feature of Blender!
Hot tip: Hitting F1 on most sections on the properties tab will open up the Blender Reference Manual for that specific topic.
#8 - Lighting & Compositing
Following immediately after the modeling stages, I took a bit of time working on lighting the environment. Initially, I wanted to emulate the lighting from the source material but found that I wanted to emphasize the light generated from the lantern to build that indoor/underground atmosphere.
I realized, realistically, it didn't seem to make sense that a single source could fill an entire scene up with light (then again, realism isn't the reason why we enjoy Miyazaki's films in the first place).
Image: From the source material, it was as if certain flora were also emitting some glow. This was later removed.
Image: A more faithful recreation of lighting in the scene.
Image: Notice how there appears to be areas lit up despite being far away from the physical light source (the lantern). It was a balancing act between realistically portraying lighting and ensuring color detail was not lost in the darkest parts of the scene.
I believe that color and lighting are pivotal in moving the audience's emotions in a certain direction. Studio Ghibli is notable for using color to enhance a narrative (thanks in part to Michiyo Yasuda, RIP).
I wanted to evoke a sense of calm and nostalgia to this scene, hence why I decided for a more atmospheric approach to lighting and colors.
#9 - Rendering!
The last challenge I faced when creating this scene was handling render artefacts. These artefacts came in the form of fireflies, noise and even de-noised pixels. You can do a Google search to find out about these artefacts as it can get very technical.
Denoising was a solution but only for image renders. When it came to rendering animation, I need to find a way to reduce the fireflies present in my scene. Despite my research online I was never really able to decrease the amount of fireflies in my render.
It was by pure luck that choosing to switch from CPU-GPU (hybrid) rendering to GPU-only rendering, significantly reduced the presence of the firefly artefacts. My hypothesis to this is that my CPU was not as capable in rendering the samples quite as my GPU and, therefore would often render pixel values as brighter than needed. Again, I'm no expert so take this with a grain of salt.
Image: Pixels overblown with light are known as fireflies.
Image: note the bright, overblown pixels surrounding the scene and flickering from the hair particles in the spore assets which were a result of the roughness value in the hair material (being slightly glossy).
Image: Composite nodes used for the different styles of the final render.
Image: Mist pass used compositor to control color in the scene depth.
Image: Raw vs Final render (denoised, 1024 samples).
#10 Conclusion (lasting thoughts)
Looking back on it now, there are elements I'd like to improve upon: deeper color saturation for and variation, soft yet rich focal length, detailed texturing and larger volume of flora. When it comes time to revisiting this style of art, I will most definitely take these elements into consideration!
To those who have continued reading thus far, thank you! I've invested in quite some time to making this post in order to inspire others to push onward with their creative career. I'll finish this off with some parting thoughts...
Dailies - Take screenshots of what you're working on. Not only are dailies used in the industry, you never know when you want to collate them into a clip or gif and show others your progression!
Perseverance - There will be moments you're not too sure about yourself or mistakes that set you back. However, that's all they really are... Setbacks. For personal projects, you can always tweak or redo parts of your project until you're happy with what you got. If you're running on a deadline, it is better to seek alternative solutions or find a middle-ground than brute-force a certain method. You can also ask for critique in the right platforms. The most important thing is that you learned something along the way.
Have Fun! - The journey you're undertaking is a creative one, so enjoy yourself! Experiment, play around, innovate. You'll never know unless you try and like I mentioned before, it isn't wasteful if you've learned something after all. Who knows, you might create something entirely awesome!
Hopefully, this has been helpful or informative and I am looking forward to what the future will bring in 3D as software and systems become more intuitive and user-friendly! Keep on keeping on!
Comments