Humanization

Talos routes aim, timing and movement through a composable humanization layer. There are two independent dials: a stationary profile (raw/natural/paranoid) and Human mode (a session-arc fatigue model plus eased cube-aim).

Note: This is best-effort obfuscation of long-session statistical detection, not a guarantee of undetectability, and automation may still violate a server's rules. It models only motor-level imperfection (overshoot, hesitation, breaks) — never semantic mistakes like attacking the wrong target.

Profiles

set_profile("raw" | "natural" | "paranoid") (Python) or the profile Talos reports on /talos human off. Profiles are categories of distributions, not fixed knobs — the trajectory family varies (Bézier / minimum-jerk / piecewise-noise), because a fixed distribution shape is itself a fingerprint.

• raw — minimal shaping; fastest.
• natural — right-skewed timing, bounded velocity/accel, overshoot + micro-correction.
• paranoid — widest variance, most conservative motion.

set_seed(seed) makes a run reproducible (one SeededRng per task, never a global) — useful for testing.

Tuning — more/less humanisation, per-knob

The profiles are starting points, not the ceiling. Two layers of user tuning sit on top, settable from Python or chat and persisted in the mod config across sessions:

Intensity — one dial

talos.intensity(1.5) or /talos human intensity 1.5. Scales the humanness knobs together: reaction delays, overshoot probability/magnitude, timing jitter and path wobble scale up with intensity, while rotation speed scales down. 0 is near-robotic, 1 is the profile as authored, 3 is the exaggerated maximum.

Per-knob overrides

talos.tune(overshoot_prob=0.3, rotation_speed_max=12) or /talos human set <knob> <value> (knob names tab-complete). Every knob is clamped into a safe range — bad values can tune aim but never break it:

Trajectory families

Restrict which aim-path shapes are used: talos.tune(families=["bezier", "min_jerk"]) — options bezier, min_jerk, linear.

Inspect & reset

talos.human_knobs() returns a dict with profile, intensity, human_mode, families, overrides (your tuning) and effective (the final numbers actually used for aim/timing); /talos human show prints the same in chat. Clear everything with talos.reset_tuning() or /talos human reset.

Design note: knobs, families and intensity are the supported way to change how humanisation behaves. Python callbacks cannot supply aim curves directly — aim plans are computed on the game thread, and the game thread never enters Python (a core stability invariant).

Human mode — /talos human [on|off] · talos.human(True/False)

The single Human-mode toggle. On bundles two things:

1. Eased cube-aim (no direct snap)

Command and Python aim — absolute angles, coordinates, blocks and entities — runs through the cube-aim controller (AimController):

• A 1×1m yellow guide cube is rendered off-grid, centred exactly on the intended point.
• The actual aim spot — the red X — lands on a visible face chosen with probability proportional to that face's visible flat area, center-biased, then a random spot on that face.
• Rotation draws a random fast sensitivity far out and a random slow sensitivity for the final approach, blending smoothly (never instantly) once the look ray passes within 0.4m of the cube, with per-tick jitter and a red dotted preview line tracing the exact curve the crosshair is about to draw.
• A per-session quadratic speed modulation (peaking mid-flight, never linear) adds a small speed swell/sag and a slight bow to the path.

/talos track keeps the same aim session alive against a moving target. Navigation gaze during /talos goto speaks the same language against a full-block cube on the look-ahead target's hitbox — while the walking bearing stays honest (the mark pulls the gaze only a few degrees off the route line; the physics rollouts still choose the actual movement inputs).

2. Session-arc fatigue

On top of the stationary profile, a wall-clock fatigue model (SessionArc) drifts your behaviour over the session so the input stream is non-stationary — because a real human's parameters drift, and a fixed-parameter profile is exactly what a server can find over hours:

• reactions slow and spread,
• aim loosens and overshoots more,
• the walk wobbles wider,
• idle micro-breaks pause pathing briefly — more often, and longer, as fatigue rises.

The HUD shows a human » fatigue N% (Mm) line (or on break (Ns)) while it's on.

Off uses direct snap aiming and disables the session drift.

Python surface

talos.human(True)        # enable; talos.human() returns current state
talos.human(False)       # disable
talos.fatigue()          # 0.0–1.0 current fatigue
talos.on_break()         # True while an idle micro-break is pausing pathing
talos.set_profile("natural")
talos.set_seed(1234)
talos.wait_between(2, 5)  # right-skewed humanized pause (the human-ish sleep)