Basically, Blender says "start with a cube". I want to ask why and what are the other options.
As far as workflows go, there are far too many to list, and most artists use multiple, but common ones are:
* Sculpt from a base mesh, increasing resolution as you go with remeshing, subdivision, dyntopo, etc. * Constructively model with primitives and boolean modifiers. * Constructively model with metaballs. * Do everything via extrusion and joining, as well as other basic face and edge-oriented operators. * Use geometry nodes and/or shaders to programmatically end up with the result you want.
On the other end of the spectrum you have "sculping" for organic shapes and you might want dig into the community around ZBrush, if you want to fill the gap from "start with a cube" to real sculpting.
More niche, but where I feel home, is the model with coordinates and text input approach. I don't know if it has a real name, but here istwhere is pops up and where I worked with it:
- In original AutoCAD (think 90s) a lot of the heavy lifting was done on an integrated command line hacking in many coordinates
- POVRay (and its predecessors and relatives) has a really nice DSL to do CSG with algebraic hypersurfaces. Sounds way more scary than it is. Mark Shuttleworth used it on his space trip to make a render for example.
- Professionally I worked for a couple of years with a tool called FEMAP. I used a lot of coordinate entry and formulas to define stuff and FEMAP is well suited for that.
- Finally OpenSCAD for a contemporary example of the approach.
- sculpting. You start with a high density mesh and drag the points like clay.
- poly modeling. You start with a plane and extend the edges until you make the topology you want.
- box modelling. You start with a cube or other primitive shape and extend the faces until you make the shape you want.
- nurbs / patches : you create parts of a model by moving Bézier curves around and creating a surface between them
- B-reps / parametric / csg / cad / sdf / BSP : these are all similar in that you start with mathematically defined shapes and then combine or subtract them from each other to get what you want. They’re all different in implementation though
- photogrammetry / scans : you take imagery from multiple angles and correlate points between them to create a mesh
My main tool is OpenSCAD, as I am a programmer and I model for 3D printing, which support both CSG and extrusion, but not the mesh manipulation like Blender.
One paradigm is not better than the other. Blender is nice for modelling, for example, characters, but it's really painful for doing precise shapes. For modelling multi-part action figures for 3D printing you'll need both things, sometimes modelling something in a tool and finishing it in other
Other popular approaches to designing shapes are using live addition/subtraction, Nomad supports that as well, but it's not as intentional as say 3D Studio Max (for Windows) - where animating those interactions is purposely included as their point of difference.
There's also solid geometry modelling, this is intended for items that would be manufactured, for that SolidWorks is common.
Then finally you have software like Maya which lets you take all manner of approaches, including a few I haven't listed here (such as scripted or procedural modelling.) The disadvantage here is that the learning curve is a mountain.
And if you like a bit of history, PovRay is still around where you feed it text mark-up to produce a render.
You can create whatever start up file you want.
Other approaches include subsurface workflows, metaballs, signed distance functions, procedural (geometry nodes) and grease pencil.