Differential Mesh Growth in Cinema 4D Scene Nodes: Ruffled Surfaces Step by Step (2026)
Ruffled, leaf-like surfaces can be grown in Cinema 4D Scene Nodes by repeating one loop inside a Memory node: select the border, subdivide and triangulate it, push the border points with noise, relax them, then remesh and smooth.
- Start: a Disc node, outer radius 25 cm, 3 rotation segments; Constant Shading (Lines) to watch the mesh.
- Select: Border Selection, Grow Selection (Edges, Steps input) and Triangulate, grouped in a Scaffold.
- Divide: Transfer Selection to polygons, Polygons Info, distance from centre to vertices averaged with Aggregate; a Loop-Carried Value subdivides only where the average passes a Divide Factor.
- Grow: a Point Modifier with Sample Noise and Range Mapper scales the normal into the position; Relax Points with Hash-random steps.
- Finish: Retriangulate and Smooth Geometry (2 iterations, 50%); after the Memory node, a Loop Subdivide and a Smoothing Deformer.
Differential growth is a biological effect: when one part of a tissue grows faster than the rest, the surface cannot stay flat and buckles into folds and ruffles, like the edge of a lettuce leaf or a flower petal. Nervous System’s Floraform project and many artists have explored it digitally.
This tutorial simulates staged surface growth in Cinema 4D Scene Nodes, starting from a single triangle. The guide follows the video chapter by chapter.
Checked against the Maxon Cinema 4D help pages for Scene Nodes and the Node Editor, November 2026.
The full tutorial runs 21:48. Chapters: 0:52 initial mesh and algorithm, 3:14 Memory solver node, 3:27 border selection, grow selection and triangulate, 4:06 divide with loop subdivide, 9:12 repulsive and spring forces, 14:10 remesh and smoothing, 16:01 second subdivide and smoothing deformer, 21:08 wrap-up.
The growth loop
The algorithm is short and runs once per frame. Select the border edges. Add points to them by subdividing, and triangulate the surface. Push those points by a random amount and relax them so they spread evenly. Remesh and smooth. Repeat. Because new points are only added at the edge, the rim gets longer every frame while the middle stays the same size.
How to read the figure. The loop is computed in 2D on a fanned disc so each step is easy to see: the triangle’s border is split, its points pushed outward by uneven amounts, then relaxed. After several loops the border has many more points and has started to wave.
1. Initial mesh
- Start with a Node Mesh.
- Add a Disc node with Outer Radius 25 cm and Rotation Segments 3, so the growth starts from the simplest triangle.
- Set the Shading Mode to Constant Shading (Lines) so every change to the mesh is visible.
2. Memory node and border selection
- Add a Memory node and connect the initial mesh to it. All the steps below run inside it, so each frame starts from the previous result.
- Use Border Selection with the active selection set to Border, then Grow Selection with Type set to Edges and an input on its Steps port, default 1.
- Add Triangulate with Type set to Polygons and the default selection string.
- Group these nodes in a Scaffold to keep the graph tidy.
3. Divide: subdivide only large border polygons
- Convert the selection from edges to polygons with Transfer Selection.
- Add Polygons Info with its selection string set to Active, and Iterate Collection on its Vertex Indexes port. Use Get Element nodes to read each polygon’s index and centre, and Get Positions with another Iterate Collection and Get Element to read its corner points.
- Measure the Distance from the centre to each corner and average it with Aggregate in Stream mode, Operation Average.
- Build an integer array of polygon indices with Build Array.
- In a Loop-Carried Value scoped to the Iterate Collection, add a Subdivide node with a Subdivisions input (default 1). A Compare node set to Greater Than or Equal checks the average distance against a Divide Factor (1), and an If node (GeometryObject) only uses the subdivided mesh when the test passes.
- Expose the Subdivide mode as an input. In the video, Catmull-Clark gave the best edge subdivision. Group this as a Scaffold named Divide – Loop Subdivide.
How to read the figure. The top row only measures: for each border polygon it averages the distance from its centre to its corners, which is a simple way to say “how big is this polygon”. The If node in the second row then uses that answer to keep either the subdivided or the original polygon. The strip at the bottom shows the effect: small polygons are left alone, large ones are split, so the border keeps an even density as it grows. If nothing ever subdivides, lower the Divide Factor.
4. Repulsive and spring forces
- Add a Point Modifier. Inside, connect a Sample Noise node to the Position port and control its intensity with a Range Mapper.
- Multiply with a Scale node and a Point step input, use the result to scale the Normal, and feed it through an Add node (vector) back into Position. The border points start to shift.
- Expose controls for the noise.
- Add Get Positions, Relax Points and Set Positions, with a Hash node to randomise the relax steps. Expose the relax controls; the video sets the relax step multiplier to 1.25.
5. Remesh, smooth and refine
- Add Retriangulate, then Smooth Geometry with Iterations 2 and Strength 50%.
- Try Loop Subdivide at 2 and Relax Strength around 17%. The surface now grows in stages: differential mesh growth.
- Add a Smoothing Deformer for a softer surface.
- After the Memory node, insert another Subdivide set to Loop with more subdivisions to refine the border edges.
- Optionally finish with a Subdivision Surface generator of type OpenSubdiv Loop.
How to read the figure. These are computed shapes, not frames from the simulation. Each disc has the same radius, but its rim is longer than a circle of that radius: 3% longer gives gentle waves, 38% longer gives tight ruffles. That extra rim length is exactly what the loop adds every frame, which is why the surface ruffles instead of staying flat.
The video notes that this method does not fully solve self-intersections and collisions, so check the mesh before rendering and bake it once you like the result.
Render the animation on several GPUs
Procedural setups are evaluated before each frame renders, so time a test frame first and cache anything that simulates over time. After that, Redshift can render each frame on every GPU in the machine. Each card holds its own copy of the scene, so plan VRAM per card, not in total (Maxon support). A scene that fits on one RTX 4090 (24 GB) or RTX 5090 (32 GB) fits on each card of a 4x or 8x server.
Differential mesh growth checklist
Check these points before you bake and render.
SaaS or IaaS render farm for procedural Cinema 4D setups?
A Scene Nodes or particle setup is stored in the .c4d file, so any farm with a compatible Cinema 4D version renders it the same way. The differences are version matching, how heavy the evaluation is before each frame renders, and how fast you can fix a setup that behaves differently at render time.
SaaS or IaaS for procedural scenes
How each farm model fits a procedural Cinema 4D scene built with nodes, fields or simulations.
| Your situation | SaaS render farm | IaaS render farm (iRender) |
|---|---|---|
| The scene is final and the farm supports your exact Cinema 4D version | A good fit: upload the scene, the farm renders the frames | Works too; you open the scene and render it yourself |
| The setup depends on a recent Cinema 4D release | Check that the farm runs the same release, since node assets can change between versions | Install or update to the release you used, so the setup evaluates the same way |
| A frame looks different from your viewport | Fix it locally, upload again and resubmit | Open the scene on the server and fix it where it renders |
| You have a deadline and need to know when rendering starts and ends | Jobs join the farm’s queue, so the start time depends on how busy the farm is; many farms sell higher-priority plans to start sooner | The server runs only your job from the moment it boots, so you control when rendering starts and can time the whole job |
Simple rule: if the scene is final and the farm runs your Cinema 4D version, a SaaS farm saves you the machine work. If you need a specific release, or want to check the setup where it renders, a server you control is the safer choice.
A test workflow on an iRender server
- Upload first. Put the .c4d file and its assets on iRender Drive before you start a server; transfer and storage are free.
- Match the version. Cinema 4D and Redshift are pre-installed; check the version and update it to the one you built the scene in if needed. Your server image keeps the change.
- Check the setup. Scrub the timeline and confirm the animation evaluates the same way as on your PC; cache any simulation.
- Render a test frame. Check the Redshift log for all GPUs and the detail at final resolution.
- Render the shot. Save frames to iRender Drive and shut the server down when the job is done; multi-GPU servers are billed per minute.
Frequently asked questions
1. What is differential mesh growth?
A simulation where a surface grows faster at its edge than in its middle, so it buckles into ruffles and folds like a leaf.
2. Why start from a 3-segment disc?
It is the simplest triangle, so the growth is easy to follow from the first frame.
3. Why use a Memory node?
So each frame builds on the previous frame’s mesh and the growth accumulates.
4. How does the setup decide what to subdivide?
It averages the distance from each border polygon’s centre to its corners and subdivides only when that average reaches the Divide Factor.
5. Which subdivision mode works best?
In the video, Catmull-Clark subdivided the edges best inside the loop; a Loop subdivide is added after the Memory node to refine the result.
6. Does the method prevent self-intersections?
Not fully. The video says self-intersections and collisions are not completely solved.
7. How is the surface smoothed?
Retriangulate and Smooth Geometry inside the loop, plus a Smoothing Deformer and optional OpenSubdiv Loop at the end.
Render your Cinema 4D and Redshift projects on iRender
iRender is an IaaS GPU render farm: you rent a bare-metal server with 1, 2, 4 or 8 NVIDIA RTX 4090 (24 GB) or RTX 5090 (32 GB) GPUs, an AMD Ryzen™ Threadripper™ PRO CPU, 256 GB ECC RAM and a 2 TB NVMe SSD, connect by remote desktop and work on it like your own workstation, with full administrator access. Cinema 4D and Redshift are pre-installed, and your server image is saved between sessions.
Multi-GPU servers are billed per minute (1-GPU servers have a one-hour minimum), and sessions of 3 hours or more get 10% off. Our 4-GPU and 8-GPU servers are liquid-cooled; in our own measurements the GPUs hold 60–70°C under sustained render load. Elite support is available 7 days a week by hotline, live chat, Skype, WhatsApp and email.
Two recorded Redshift runs on iRender servers:
New users get a 100% bonus on the first top-up (minimum $50). See plans on the Cinema 4D and Redshift render farm page or download the iRender GPU app.




