M.E.C.H.A. (Mechanical Character Assembler) is a toolkit I am building for procedural assembly and rigging of mechanical characters in Houdini, using SOPs and APEX.
When looking at Hard Surface models and workflows in movies for robots and toy mechs, a huge problem was A) the lack of rigging and animation tools to test the movement and motion unless a big effort was made to export back and forth, B) Problems with rigging tools not working with extensive constraints, or needing lots of workarounds, C) the limit of scalability. 10k Parts and the rigging tool in Maya crashes, D) Model-Rigging Workflow is extremely iterative, and change-heavy, which breaks rigs and animations, E) APEX is still very fickle to me and has its usability issues, and the workflow of manually building and binding skeletons was exhausting and with extensive up and downs and back and forth with errors.
The core idea of the toolkit is mimicking real-life Gundam Modelling Kits into a tool to combine a modelling with a rigging workflow for hard surface models. You model with a Kitbash or your own model, and the toolkit handles skeleton building, rig construction, logic, and constraints based on presets.
I am posting here to get eyes on the approach and hear from anyone working in a similar territory.
Included are screenshots and short demo clips to show the current state. And I will split up the tool explanation, workflow and my questions in the following posts to keep these short.
Current results (all of these took me 10-30 minutes, including animation/posing):
Full Version: KitBash Assembler and APEX Rig Builder for Robots
WHAT IT IS
MECHA is a node-based toolkit for assembling and rigging mechanical characters directly inside a GEO context with APEX running in the background. It is built for artists who want to prototype mechanical characters (MECHA, Robots, Transformers) with motion.
The technical concept: each node is one part. You connect nodes top to bottom, and that arrangement becomes your rig. You see the result in the viewport as you build, and each part builds the skeleton and the rig at the same time. Imagine it like a Gundam Model Kit! Connecting the arm joint to the upper arm immediately allows you to move those parts.
NODES
PRIM_Part (as seen in the image above) is the base unit. It handles geometry and model parts, transforms, skeleton building, and rig settings, including single-part constraints and controls.
PRIM_Assembly combines multiple PRIM_Part nodes into a single rigged structure, such as Arm, Leg, Spine, etc., and the assembly handles the rig construction. Tags and names of the parts determine how they are treated. Alternatively, the APEX Autorig Builder is also compatible with this toolkit. ASSEMBLY also has group utility operations, such as mirror and transform, that apply to geometry, skeleton, and rig.
MECHA BUILDER lets you import a finished model to input it into the tool, with automatic tagging based on part naming and carries over the hierarchy made in Maya or Blender for quick skeleton build and rigging.
Below, I have also added images of some of the UI and a custom Python Panel Kitbash Library that supports the tool.
Supporting Nodes:
PISTON takes two inputs, upper and lower piston, and handles the look-at and stretch behaviour between them automatically.
MERGE_BRANCHES combines separate assembly branches (left arm, right arm, legs) into a single output for the final rig.
TRACK is a curve constraint that constructs the geometry at the same time. Parts multiply and distribute along the line. Used for tank treads, chainsaw chains, and sliding armour panels.
KITBASH CONVERT lets you input your own kitbash set to use in the Houdini internal toolkit Kitbash Library.
WORKFLOW
Two paths in. (see videos below)
1. You either import a pre-modelled FBX and use the BUILDER that automatically builds the tool for you, and then you fix the hierarchy of nodes for the rig, or
2. You build directly from the KitBash library using the the PART nodes and the viewport tools. You can exchange parts and extend the model at any time without the rig breaking.
MECHA is a node-based toolkit for assembling and rigging mechanical characters directly inside a GEO context with APEX running in the background. It is built for artists who want to prototype mechanical characters (MECHA, Robots, Transformers) with motion.
The technical concept: each node is one part. You connect nodes top to bottom, and that arrangement becomes your rig. You see the result in the viewport as you build, and each part builds the skeleton and the rig at the same time. Imagine it like a Gundam Model Kit! Connecting the arm joint to the upper arm immediately allows you to move those parts.
NODES
PRIM_Part (as seen in the image above) is the base unit. It handles geometry and model parts, transforms, skeleton building, and rig settings, including single-part constraints and controls.
PRIM_Assembly combines multiple PRIM_Part nodes into a single rigged structure, such as Arm, Leg, Spine, etc., and the assembly handles the rig construction. Tags and names of the parts determine how they are treated. Alternatively, the APEX Autorig Builder is also compatible with this toolkit. ASSEMBLY also has group utility operations, such as mirror and transform, that apply to geometry, skeleton, and rig.
MECHA BUILDER lets you import a finished model to input it into the tool, with automatic tagging based on part naming and carries over the hierarchy made in Maya or Blender for quick skeleton build and rigging.
Below, I have also added images of some of the UI and a custom Python Panel Kitbash Library that supports the tool.
Supporting Nodes:
PISTON takes two inputs, upper and lower piston, and handles the look-at and stretch behaviour between them automatically.
MERGE_BRANCHES combines separate assembly branches (left arm, right arm, legs) into a single output for the final rig.
TRACK is a curve constraint that constructs the geometry at the same time. Parts multiply and distribute along the line. Used for tank treads, chainsaw chains, and sliding armour panels.
KITBASH CONVERT lets you input your own kitbash set to use in the Houdini internal toolkit Kitbash Library.
WORKFLOW
Two paths in. (see videos below)
1. You either import a pre-modelled FBX and use the BUILDER that automatically builds the tool for you, and then you fix the hierarchy of nodes for the rig, or
2. You build directly from the KitBash library using the the PART nodes and the viewport tools. You can exchange parts and extend the model at any time without the rig breaking.
To improve the system, I am looking for input on the following items. It would be helpful if you can find time and reply with your thoughts on any of these. If it works better, I can also set up a chat or meeting, just let me know in a DM.
WHAT I AM LOOKING FOR:
On scalability: has anyone hit real limits with node-based SOP approaches at part counts above a few hundred? I have not stress-tested this at the 10k range. I want to know what breaks. But I have previously seen Michael-Bay-level transformers used with APEX.
On the iterative model-rig workflow: the toolkit keeps geometry and rig in the same node graph so model changes do not break the rig. Does this match problems you have actually run into? What would a real iteration cycle look like for your work?
On APEX: If you have worked with APEX, where does this toolkit add value, or work with your existing workflow, and where does it create friction compared to your existing approach? I am not trying to replace APEX for production rigs. I want to know where the boundary should be.
On SOPs Friction: One issue I am struggling with is how to make it more compatible with current Houdini SOPs Nodes, as the toolkit is essentially all nodes packaged into APEX. Meaning that the tool does not allow for simple transform or other SOP nodes to be used. I would love a perspective on the necessity of this.
On constraints: Are there constraint setups for mechanical characters that this approach would not handle? Pistons, look-ats, locks, parenting, tracks, and basic IK are covered. What is not covered that you would need?
And I welcome any other thoughts, questions, criticisms, or tips!
WHAT I AM LOOKING FOR:
On scalability: has anyone hit real limits with node-based SOP approaches at part counts above a few hundred? I have not stress-tested this at the 10k range. I want to know what breaks. But I have previously seen Michael-Bay-level transformers used with APEX.
On the iterative model-rig workflow: the toolkit keeps geometry and rig in the same node graph so model changes do not break the rig. Does this match problems you have actually run into? What would a real iteration cycle look like for your work?
On APEX: If you have worked with APEX, where does this toolkit add value, or work with your existing workflow, and where does it create friction compared to your existing approach? I am not trying to replace APEX for production rigs. I want to know where the boundary should be.
On SOPs Friction: One issue I am struggling with is how to make it more compatible with current Houdini SOPs Nodes, as the toolkit is essentially all nodes packaged into APEX. Meaning that the tool does not allow for simple transform or other SOP nodes to be used. I would love a perspective on the necessity of this.
On constraints: Are there constraint setups for mechanical characters that this approach would not handle? Pistons, look-ats, locks, parenting, tracks, and basic IK are covered. What is not covered that you would need?
And I welcome any other thoughts, questions, criticisms, or tips!