Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ASTRA228b/Experimental/llms.txt

Use this file to discover all available pages before exploring further.

APreds (Astra Predictions) reads the instantaneous velocity of each hand via a GorillaVelocityTracker and nudges the hand’s position forward along its movement direction before the frame renders. The result is a subtle extension of each reach that can help you tag surfaces or players that are just outside your normal arm range — especially during fast swings or launches.

​
How the Tracker Works

When APreds is enabled, the mod creates two invisible GameObject cubes (one per hand) and attaches a GorillaVelocityTracker component to each. Every frame these objects are repositioned to match the current left and right hand positions so the tracker can sample hand velocity. When APreds is disabled the objects are destroyed, leaving no persistent scene changes. 
EnablePreds  → Instantiate cube → attach GorillaVelocityTracker
MakePreds    → reposition cube to hand position each frame → sample velocity
DisablePreds → Destroy tracker GameObjects
Because the tracker objects are created at runtime and destroyed on disable, toggling APreds mid-session is safe and does not accumulate stale objects in the scene hierarchy.

​
Parameters

Type: float | Range: 0.001 – 0.200The scalar applied to the hand’s velocity when computing the predicted target position:
target = hand.position + velocity * PredSrength;
Low values (0.01–0.05) produce a barely noticeable extension, while the maximum (0.20) pushes hands significantly ahead of their physical position. Start low and increase gradually — very high values can make hands feel detached.
Type: float | Range: 0.01 – 0.300The minimum velocity magnitude (in units/frame) required before prediction is applied to a hand. If the hand velocity’s magnitude is below this value, the hand position is left untouched. Raise this to prevent micro-jitter prediction when your hands are nearly still; lower it for prediction that activates even during slow reaches.
Type: float | Range: 0.01 – 0.500The t value passed to Lerp when blending from the current hand position toward the predicted target:
hand.position = Vector3.Lerp(hand.position, target, smoothness);
Lower values (0.01–0.1) produce gradual, floaty prediction. Higher values (0.4–0.5) snap hands quickly to the predicted position. Match this to your movement speed — faster movement generally benefits from higher smoothness.
Type: float | Range: 1.0 – 2.5After the prediction lerp is applied, each hand is clamped to at most maxArmLength units from the player’s head position. This prevents extreme prediction values from stretching your arms to physically impossible lengths or clipping through distant geometry.

​
Presets

PresetPredSrengthNotes
Max0.200Full prediction — hands lead movement significantly
Random0.001 – 0.200Randomises prediction strength within range
Reset0.001Near-zero prediction, essentially off
There is no dedicated “Legit” preset for APreds. Build your own by starting at PredSrength = 0.030, smoothness = 0.15, and movementThreshold = 0.05, then adjust from there.

​
Getting Started

1

Open the APreds window

Launch Experimental and open the E - APreds UI window from the mod menu.
2

Enable the mod

Toggle IsPredOn on. The two tracker GameObject cubes are created automatically.
3

Set prediction strength

Begin with a low PredSrength value (0.02–0.05) and swing your arms to feel the extension. Increase incrementally until reaches feel satisfying.
4

Tune the threshold

If you notice prediction firing when your hands are resting, raise movementThreshold until jitter-prediction stops.
5

Adjust smoothness and arm length

Lower smoothness for a fluid feel or raise it for snappier response. Reduce maxArmLength if your hands appear to reach through walls.
Setting maxArmLength above 2.0 may allow hands to intersect distant surfaces or players unintentionally. Keep it at or below 1.8 unless you specifically need extended reach.

Build docs developers (and LLMs) love

Get started for freeTalk to us