pGradientForce

pGradientForce is a Fusion particle force node that uses the spatial gradient of the alpha channel in an input 2D image as a vector field. It does not treat RGB as an ordinary picture or convert the image into particles; the Image input supplies the map whose alpha changes accelerate an existing particle stream.

The node has two permanent inputs. The orange Particles input receives a stream from pEmitter, pImageEmitter, pMerge or another particle node. The green Image input receives a 2D image such as Fast Noise, a masked Background or a prepared map, and this input is what distinguishes pGradientForce from ordinary one-field forces.

A useful mental model is terrain. Particles react to spatial change in alpha rather than absolute alpha alone: where alpha changes, the node can derive a force direction. A flat alpha plate with no slope provides little useful gradient direction even if its RGB looks detailed.

With positive Strength, particles accelerate along the alpha gradient from high values toward low values — white to black. The reference manual describes this as giving particles the appearance of moving downhill or following the contour of a provided shape.

Negative Strength does not merely weaken the effect. It reverses the Gradient Force so particles accelerate from low alpha toward high alpha, black to white. The sign of Strength is therefore a genuine direction control.

Strength controls force and acceleration, not a fixed particle speed. The force contributes to the simulation over time, so its result accumulates. Excessive absolute Strength can throw particles out of the useful map area quickly and make the field difficult to read.

Fast Noise is a useful diagnostic source because it can provide continuous alpha transitions and immediately reveal non-uniform steering. Production maps can be radial ramps, silhouettes, text mattes, distance-like fields or animated alpha textures.

Keep Gradient Image and Region conceptually separate. The green Image input determines direction and force through its alpha gradient. Region only decides where pGradientForce is allowed to act. Supplying an image only as a Bitmap Region does not replace the main gradient Image input.

By default there is no separate Region socket beyond the two permanent Particles and Image inputs. A third Region input appears dynamically only when Region is set to Bitmap or Mesh.

Built-in Regions such as All, Bezier, Cube, Line, Rectangle and Sphere are generated inside the node and need no cable. Bitmap Region accepts another 2D image/channel while Mesh Region accepts 3D geometry; both are independent of the permanent gradient Image input.

Conditions > Probability controls the chance that an eligible particle is processed on each frame. It is not one permanently selected percentage for an entire lifespan; values below 1 are evaluated for each particle on each frame.

Age Start/End defines a normalized portion of particle lifespan rather than timeline frame numbers. A range such as 0.3–0.8 can leave launch motion alone, enable gradient steering through the middle of life and disable it again near death.

Set Mode and Set 1–32 filter particle Sets that were assigned upstream. pGradientForce uses those Sets for eligibility; it does not create Set membership by itself.

Region Mode defines the spatial relationship — ignore, intersecting, inside, not intersecting, not inside, entering or leaving — while Region chooses geometry. Neither replaces the gradient Image source.

Bitmap Region is useful when one image defines force direction and another image defines where that force is permitted. In that topology the node legitimately has both its green gradient Image and a separate dynamic Bitmap Region socket.

Mesh Region adds a 3D spatial gate with common particle Region controls for surface/volume interpretation, winding and ObjectID. The vector field still comes from the 2D Image input; the mesh does not become the gradient source.

Neighboring particle forces should be introduced one at a time. pGradientForce changes Velocity, pFriction can damp the gained speed, pTurbulence can add noise, and other forces can sum with the motion. Adding all of them at once hides causality.

pGradientForce does not create RGBA output. Its output remains Particles and must reach pRender before ordinary image tools such as Merge, Glow, ColorCorrect and MediaOut can be used.

A strong diagnostic graph is FastNoise → Image plus pEmitter → Particles into pGradientForce, then pRender. Start with Probability = 1, full Age, Ignore Sets and Ignore region, and change only Strength and the alpha gradient to prove white→black and black→white behavior.

If particles barely react, inspect the source alpha before increasing Strength. A common failure is an image with strong RGB contrast but nearly constant alpha, leaving pGradientForce with a flat field.

If motion feels unpredictable, replace noisy input with a static smooth gradient temporarily. Once the sign and direction are proven on a simple map, restore Fast Noise or an animated matte.

In production, treat pGradientForce as image-driven steering: design the field independently from emission. This allows direction to vary across the frame and follow alpha contours instead of applying one global direction like a simple directional force.