Why Blender Is the Best Software for the 3D Workflow
The 3D workflow has grown over the decades and boasts hundreds of different software tools. In this article, we talk about the main parts of the workflow, what tools exist for each, and why we think Blender is the best tool of them all (especially for synthetic data).
3D, short for the three dimensions of space we live in, is a catch-all term used to describe the varied technologies used to create virtual worlds. 3D’s technology stack can be roughly split into two broad categories: asset creation and asset scripting. Asset creation is the process of creating assets: virtual objects, scenes, and materials. Asset scripting is the process of manipulating those assets and their interactions over the fourth dimension of time. Decades of progress have resulted in sophisticated software tools that make 3D workflows more automated and straightforward, but a significant amount of human expertise and artistic talent is still required.
Assets are digital representations of a 3D object. One type of asset is a mesh: a connected graph of 3D points also called vertices, which define the surface of an object. Edges interconnect vertices, and a closed loop of vertices creates a polygon known as a face. The engineering and manufacturing world creates meshes using computer-aided design (CAD) software such as AutoCAD, Solidworks, Onshape, and Rhino. The entertainment industry creates meshes using modeling software such as Maya, 3DSMax, and Cinema4D.
Whereas a mesh describes the shape and form of an object, a material asset describes the texture and appearance of a virtual object. A material may define rules for the reflectivity, specularity, and metallic-ness of the object as a function of lighting conditions. Shader programs use materials to calculate the exact pixel values to render for each face of a mesh polygon. Modeling software usually comes packaged with tools for the creation and configuration of materials.
Finally, asset creation encompasses the process of scene composition. Assets can be organized into scenes, which may contain other unique virtual objects such as simulated lights and cameras. Deciding where to place assets, especially lights, is still almost entirely done by hand. Automatic scene composition remains a tremendous challenge in the 3D technology stack.
The fourth perceivable dimension of our reality is time. Asset scripting is the process of defining the behaviors of assets within scenes over time. One type of asset scripting is called animation, which consists of creating sequential mesh deformations that create the illusion of natural movement. Animation is a tedious manual task because an artist must define every frame; expert animators spend decades honing their digital puppeteering skills. Specialized software is often used to automate this task as much as possible, and technologies such as Motion Capture (MoCap) can be used to record the movement of real objects and play those movements back on virtual assets.
Game Engines are software tools that allow for more structured and systematic asset scripting, mostly by providing software interfaces (e.g., code) to control the virtual world. Used extensively in the video game industry after which they were named, examples include Unity, Unreal Engine, GoDot, and Roblox. These game engines support rule-based spawning, animation, and complex interactions between assets in the virtual world. Programming within game engines is a separate skillset to modeling and animating and is usually done by separate engineers within an organization.
Blender is an open-source 3D software tool initially released in 1994. It has grown steadily over the decades and has become one of the most popular 3D tools available, with a massive online community of users. Blender’s strength is in its breadth: it provides simple tools for every part of the 3D workflow, rather than specializing in a narrow slice. Organizations such as game studios have traditionally preferred specialization, having separate engineers using separate tools (such as Maya for modeling and Unreal Engine for scripting). However, the convenience of using a single tool, and the myriad advantages of a single engineer being able to see a project start to finish, make a strong case for Blender as the ultimate winner in the 3D software tools race.
Many of the world’s new 3D developers opt to get started and build their expertise in Blender for its open-source and community-emphasizing offering. This is an example of a common product flywheel: using a growing community of users to improve a product over time. With big industry support from Google, Amazon, and even Unreal, Blender also has the funding required to improve its tools with this user feedback.
In addition to supporting the full breadth of the 3D workflow, Blender has the unique strength of using Python as the programming language of choice for asset scripting. Python has emerged as the lingua franca for modern deep learning, in part due to the popularity of open-source frameworks such as TensorFlow, PyTorch, and Scikit-Learn. Successful adoption of synthetic data will require Machine Learning Engineers to perform asset scripting, and these engineers will be much more comfortable in Blender’s Python environment than Unity’s C# or Unreal Engine’s C++ tools.
Thanks for getting this far! If you’re interested in 3D and what it can do for synthetic data, check out our open-source data development toolkit zpy. Everything you need to generate and iterate synthetic data for computer vision is available for free. Your feedback, commits, and feature requests are invaluable as we continue to build a more robust set of tools for generating synthetic data. In the meantime, if you need our support with a particularly tricky problem, please reach out.