FxForge: Realistic Destruction VFX in Blender

Blender destruction VFX looks fake for three reasons: debris at the wrong scale, dust that shows up before the impact and a camera that never reacts. FxForge has spent years breaking things on camera, so here is how a believable shot gets planned.

FxForge: the channel that breaks things and makes it look good

FxForge is the work of Albin Thorburn (@AlbinThorburn), who started the channel in November 2019 and has built it to roughly 110,000 subscribers, 61 videos and more than 7.7 million views. His specialty is the hardest thing to get right: shots where something large collapses and still reads as real two seconds later.

The channel runs two complementary formats. Tutorials first: "blender large scale destruction" (58,690 views, June 2026), "Blender car crash tutorial!" (122,491, November 2025) or "How to simulate a car crash in blender". Then breakdowns of his own scenes: "How I made this bridge destruction scene in blender" (77,801, May 2026), "Making Stranger Things vfx in blender" (103,341) or "blender dynamic vegetation" (15,366). The tutorials teach technique; the breakdowns show how a whole shot gets decided.

His biggest uploads follow that same "how I made this" logic: "Blender Monster VFX Tutorial" (537,380 views, September 2025) and "can you recreate Marvel CGI in blender?" (280,337). Finished shots, solved end to end, shortcuts and headaches included.

On this platform, FxForge is curated as 10 videos at /channels/115: nine in the Blender category and one in Maya, all tagged as beginner difficulty. You do not need to be a seasoned FX veteran to start breaking things with control. The official channel lives at youtube.com/@AlbinThorburn.

What makes Blender destruction VFX believable (and what kills it)

The solver is almost never the problem: the choices that sink believability are made before you hit play, and the Blender manual is fairly blunt about them.

Scale comes first

Rigid body simulation is unstable with small objects: the manual recommends pieces at least 20 cm in diameter. Below that, mesh collision shapes do not save you either, because they have no real thickness -they are triangles with no volume- and fast objects pass straight through. If your wall shatters into coin-sized chunks, everyone can tell it is a simulation.

Impact hierarchy

Nothing breaks at the same time. Structure fails first and debris follows: that sequence is what the eye reads as weight. You control it with breakable constraints, rigid body joints that hold until the impulse crosses a threshold (Breakable plus Threshold). To join several bodies, the Connect operator creates an empty that carries the constraint and points at both selected objects. The full setup of rigid bodies, collision shapes and fracturing is covered step by step in rigid body destruction in Blender.

Dust is half the effect

Dust sells scale better than the debris itself. A shard with no cloud around it looks like a floating cube; add dust and the same shard has mass. With Geometry Nodes and a Simulation Zone you can drive simple frame-by-frame rules, where each frame's result feeds the next one, and Delta Time keeps the speed consistent when the frame rate changes. That lightweight approach is broken down in stylized smoke and fire with Geometry Nodes.

Secondary layers

Vegetation that keeps shaking, cloth that reacts a beat after the hit, soft bodies that deform: these are the layers that give a shot life, and the ones most people skip. FxForge's route touches exactly that in "blender dynamic vegetation", and it pairs well with cloth simulation and soft bodies. Once the impact is over, the eye keeps looking for secondary motion; if there is none, the shot feels dead.

Organizing a scene with several simulations

A scene with walls, debris, dust and secondaries is never simulated in one lump: it gets organized in stages.

The Rigid Body World and its collection

In Blender, rigid bodies and their constraints only take part if they live in the collection set under Scene > Rigid Body World. That unlocks something practical: create several worlds and spread objects across collections, so you can simulate one group, bake it and move to the next without everything hanging on a single giant simulation. One world for the main wall, another for loose debris, and the scene stops being a house of cards.

Caches: bake before you judge

The cache panel defines the range where the simulation is active and lets you bake it with Bake, Bake All Dynamics or Calculate to Frame. One detail eats hours and is worth remembering: if the .blend file has not been saved, the cache is created in memory and is lost when you close it. Save first, bake after.

Stability versus cost

Substeps Per Frame raises precision, not speed, while Solver Iterations makes stacks and constraints more stable. The manual warns that too many steps -over 1000- makes the simulation less stable. Deactivation switches off resting bodies: faster and steadier, with the occasional glitch as a trade-off, and Start Deactivated wakes them by bounding box proximity. If you do not need dynamic scaling, apply scale with Ctrl-A before simulating.

From viewport to render

Bake to Keyframes (under Object > Rigid Body) turns physics into animation keys and lets you render without recalculating. With the cache already on disk, the scene can go to a cloud render farm, as covered in cloud render farms, and the video sequence editor handles the assembly without leaving Blender.

Finishing: compositing and camera

The compositor is where the layers meet: passes and AOVs, film grain, lens distortion, glare. The same destruction render can look like plastic or like cinema depending on that step (passes and AOVs guide). The camera also has to react to the impact: depth of field and motion blur change how much weight the shot carries (depth of field and motion blur). When destruction is comped over live footage, tracking and keying come into play: motion tracking and matchmoving and green screen keying.

Mistakes that keep showing up

• Keyframing a passive body without ticking Animated: the object moves on screen, but physics still believes it is at its starting position.
• Using mesh collision shapes on small or very fast objects, which pass through each other.
• Rendering before baking the simulation, or losing the cache because the file was never saved.
• Simulating the whole scene at once instead of in stages.
• Filling the frame with debris and no scale control: visual noise that screams 3D.

The FxForge videos to start with

1. Base workflow: Realistic destruction effects in blender and Wall destruction simulation in blender.
2. Scale up: Realistic house destruction in blender and advanced destruction effects in blender.
3. Full scenes: How to simulate a car crash in blender and can you recreate Marvel CGI in Blender?
4. Layers and secondaries: making a medieval battle scene in blender and blender dynamic vegetation.
5. Creature VFX: Blender Monster VFX tutorial.

Other channel spotlights on the platform are Aura Prods and Blender Guru, and the wider context of the craft is in what VFX are.

Conclusion

The gap between a toy explosion and a convincing one is not the solver, it is the planning: right scale, impact in stages, dust on time, secondary layers and a camera that reacts. Start with a short shot -one wall, three layers: debris, dust and camera- and add complexity later. The curated FxForge videos are the best place to start practicing in Blender.