Mantaflow Fluid Particles in Blender: The Settings That Actually Matter
Mantaflow changed the game for fluid simulation in Blender. The old particle-based fluid system was notorious - unpredictable, buggy, prone to producing results that looked nothing like what you set up. So when Mantaflow arrived and made fluid particles not just possible but straightforward, it felt like someone finally turned the lights on in a room you’d been stumbling around in blindfolded.
This tutorial from Blender Made Easy walks through the entire process: setting up a Mantaflow liquid particle simulation in Blender 2.82a, baking it, and rendering it with a gradient material that actually looks good. It’s beginner-friendly but covers enough ground that intermediate users will find useful tidbits they might have missed.
Setting Up the Mantaflow Domain and Flow Object
Every Mantaflow simulation needs three pieces: a domain, a flow source, and optionally an effector. The domain is the container - in this tutorial, a scaled-up default cube. Everything inside that bounding box is where your simulation happens. The flow object is the smaller cube set to liquid/geometry behavior, which is what generates the fluid. An obstacle cube acts as an effector with collision type, giving the particles something to interact with.
The domain settings are where the real work happens. Under the physics tab, you set the domain to liquid type, and from there you get access to all the parameters that control how your simulation behaves. The default settings tend to produce simulations that look like someone turned on a fire hose - way too fast, too uniform, too mechanical. Which brings us to the first critical adjustment.
Taming the Simulation: Time Scale and Narrow Bandwidth
Time scale to 0.5. Do this before anything else. Mantaflow simulations default to running too fast, and the result is fluid that looks more like an explosion than a pour. Halving the time scale gives the simulation room to breathe and lets the physics produce something organic.
Increase minimum time steps to 2 for smoother results. This is one of those settings that doesn’t dramatically change the visual output but prevents weird artifacts that only show up when you’re three hours into a render and finally notice that one frame where everything jittered.
Now the big one: narrow bandwidth. This setting is the difference between a thin surface shell of particles and a full volumetric fill. At the default low value, you get particles only along the surface of your fluid - which is fine for mesh-based rendering but useless if you want to see individual particles filling a volume. Set narrow bandwidth to around 10, and the entire simulation volume fills with particles. This is the single most important setting for particle-based fluid visualization, and it’s buried in a submenu where most people never find it.
Bump randomness up to about 0.4 while you’re in there. Without it, particles sit in perfect grid formation, which looks exactly as artificial as it sounds. A little randomness goes a long way toward making the result feel like actual fluid rather than a data visualization of a spreadsheet.
Particle Sampling, Baking, and the Crash Factor
Particle sampling controls density. Higher values mean more particles, which sounds great until Blender crashes. The tutorial drops sampling to 1 for stability, and that’s solid advice. Mantaflow fluid simulations are memory-hungry, and pushing the sampling too high is one of the fastest ways to turn a productive afternoon into a lesson in frustration.
Baking locks in your simulation data frame by frame. Set your frame range - the tutorial uses 150 frames - and hit bake. Once baked, the simulation is cached and won’t recalculate on every playback, which makes scrubbing through the timeline actually usable. Until you bake, every frame change triggers a full recalculation, and on a dense particle sim that can mean waiting 30 seconds per frame just to see what happens.
Rendering Particles with Eevee and the Object Info Workaround
Here’s where things get interesting. The tutorial uses an icosphere (subdivision level 1) as the particle render object, set via the particle system’s “render as object” mode. This gives each particle a visible geometric form rather than rendering as halos or points - which would look terrible and also defeat the purpose of a particle sim.
The clever part is the material. Particle Info nodes don’t work in Eevee. Full stop. If you’re coming from Cycles where you’d use Particle Info to drive per-particle color based on velocity or age, that entire workflow is dead in Eevee. The workaround is surprisingly elegant: use the Object Info node’s location output instead. Feed that into a Mapping node, rotate 90 degrees on the Y axis, scale Z down to 0.1, then pipe through a Gradient Texture and Color Ramp. The result is a smooth vertical gradient - dark blue at the bottom, light blue at the top - applied across all particles based on their spatial position.
It’s not per-particle variation based on simulation data, but it looks good. Really good, actually. The gradient gives the particle field visual depth and structure that a flat color could never provide. Sometimes the workaround produces better results than the “correct” method would have, and this is one of those times.
One more rendering tip from the tutorial that’s worth internalizing: always render as an image sequence, never as a movie file. If Blender crashes mid-render - and with fluid simulations, it might - you lose everything from a movie file render. An image sequence preserves every frame up to the crash. Just re-render from the next frame. It’s a small habit that saves enormous headaches.
The Bottom Line
Mantaflow fluid particles in Blender have gone from “technically possible but why would you subject yourself to that” to genuinely approachable. The key settings - narrow bandwidth at 10, time scale at 0.5, randomness at 0.4, and conservative sampling - are a solid starting template for any liquid particle sim. The Object Info gradient workaround for Eevee is a technique worth bookmarking, since Eevee’s limitations with particle data aren’t going anywhere. And rendering as an image sequence is the kind of boring, unglamorous best practice that separates people who finish projects from people who rage-quit after their third crash.
The tutorial is clear, well-paced, and doesn’t waste your time. If you’ve been avoiding fluid sims because the old system burned you, this is worth watching.
Recommended Gear for This Tutorial
| Wacom Intuos Graphics Tablet | Logitech MX Master 3S Mouse |
Watch the Tutorial
Level Up Your Blender Setup
| Mechanical Numpad for Blender | Blender For Dummies Book |
