Anime-style VFX in Blender looks like a shader problem. It is not. You can apply a perfect toon shader to a smoke simulation and it will still look like 3D pretending to be 2D.
Hiroshi Kanazawa’s explosion work is the clearest demonstration of what actually separates the two — and the answer is mostly about what you remove.
Every 3D Default Works Against You
This is the part worth internalising before touching a node. A renderer is built to simulate light and motion accurately. Anime VFX is a set of deliberate inaccuracies:
- Soft volumetrics vs hard-edged shapes. Anime smoke has crisp boundaries. Volume simulation produces soft gradients by definition.
- Physically correct falloff vs flat bands. Real light falls off smoothly. Cel shading quantises it into discrete steps.
- Smooth interpolation vs snappy timing. 3D eases between keys. Anime holds, then snaps.
- Realistic dissipation vs shapes that hold and pop. Smoke disperses gradually. Anime smoke expands, holds its silhouette, then changes.
You are not building an effect so much as overriding a renderer that keeps trying to be accurate.
Layering Beats Simulation
Kanazawa's approach separates and layers distinct visual elements rather than relying on a single simulation to produce everything.
That is the key structural decision. A simulation gives you one physically plausible result. Layering gives you art direction — the core mass, the outline, the debris and the flash each become elements you control independently.
Hand-drawn effects animation works exactly this way. An animator draws the shape they need, not the shape physics would produce.
Shapes Take a Day. Timing Takes a Week.
This is the part almost no tutorial covers, and it is where most attempts fail.
Anime effects live on timing contrast. A shape holds for two or three frames, snaps to the next, holds again. That stillness is what makes the snap read as impact — the eye needs a moment to rest before it can register a change as sudden.
3D interpolates smoothly by default. Everything eases from one value to the next, constantly, and nothing is ever still. That constant motion is what reads as CG long before anyone examines the render.
Practically: animate on 2s or 3s rather than every frame, use constant interpolation in the graph editor where you want holds, and resist smoothing every curve. A held pose is a decision, not a missing frame.
Silhouette Over Detail
With flat colour and hard edges doing the work, the outline carries the entire read. Detail inside the shape contributes almost nothing.
Test by rendering your effect as a flat black shape. If it is unreadable, no amount of shading will save it — and if it reads clearly, you are most of the way there.
Limited Values, Shifted Colour
Anime effects typically use very few tones — often three: a base, a shadow, and a hot core or highlight. Use hard stops on a colour ramp rather than a smooth gradient.
One detail borrowed from 2D artists: the shadow tone is usually hue-shifted, not just darkened. A shadow that is simply a darker version of the base reads as computed. Shifting it toward blue or violet reads as painted.
Geometry Nodes Is the Right Tool
Procedural setups suit this far better than simulation, for one reason: control. You decide where a shape goes and when it changes, rather than tuning solver parameters and hoping.
They are also reusable. A well-built explosion setup becomes a tool you apply repeatedly, which a baked simulation never does.
More procedural setups in our Geometry Nodes roundup.
The Impact Frame
Worth knowing if you are doing action: many anime impacts include a single frame that is not the characters at all — a white silhouette, an inverted flash, an abstract smear. It holds for one or two frames and is gone before you consciously register it.
It hides the transition between poses and implies force the animation never actually shows. Free UE5 tooling for this exists — see our stylised combat plugins roundup.
Where to Start
Get your shading fundamentals right first — our guide to clean anime and toon shading covers ramps, normals and outlines.
Then study reference frame by frame. Not to copy shapes, but to count how long each one holds. That number is the thing you cannot get from a tutorial, and it is the thing that makes the difference.
