The active phase is the bounded interval in which an attack can produce its gameplay effect. Kern makes one previewed lane dangerous only from 1.60–2.18 seconds, then continues the visible follow-through after collision ends. Tavi clears the live lane and re-enters during harmless motion; boundary flashes expose both active limits.

Open boss builder
Read startup and clearFrom 0.55–1.60 s Kern draws back while a dashed rectangle promises the complete lane. Tavi uses ordinary movement to place the whole body above it before the first active instant.
Separate a short live hitbox from a longer visible strike A stone boss prepares a horizontal strike while a dashed rectangular lane shows its reach. A sword-bearing player steps fully above the lane. The rectangle becomes solid and dangerous for a short fixed interval, then changes to a harmless dashed residue while the boss visibly follows through. The player crosses the old lane, attacks during recovery, and both characters reset continuously. Boss Player

From 0.55–1.60 s Kern draws back while a dashed rectangle promises the complete lane. Tavi uses ordinary movement to place the whole body above it before the first active instant.

Implementation checklist

Define startup, first active instant, last active instant, collision geometry, contact policy, follow-through, and recovery from one attack record.

Core checks

Three questions for the mechanic

  • What starts and ends the active phase?

    Store explicit first-active and last-active times or state transitions beside the attack definition. At 1.60 s, enable the lane and emit one start event; at 2.18 s, disable it and emit one end event. Decide inclusive and exclusive frame semantics, simulation ordering, hit-stop behavior, and whether slow motion changes game-time duration rather than leaving those details to animation sampling.

  • Which geometry exists during that interval?

    Use one rectangular lane from (282, 406) to (548, 454) for preview, collision, debug drawing, and tests. Measure Tavi's complete 24-unit radius against every edge. If the weapon translates or rotates during active time, sample that same authoritative shape continuously and test the swept volume instead of checking only its first and last poses.

  • What happens on contact and after cutoff?

    Declare whether one target can be hit once, once per active segment, or repeatedly on a cooldown, then store the consumed contacts with the attack instance. At cutoff, remove collision even though the follow-through remains visible. Preserve enough recovery for the successful response to change spacing or create a deliberate attack opening.

Design mistakes

  • Collision lasts for the entire attack animation

    A weapon can keep moving after its useful striking interval. If the hitbox survives through every follow-through frame, Tavi is hit while entering an opening the pose appears to promise. Give collision its own bounded state, and let the later motion communicate weight without silently extending danger.

  • Visual and gameplay boundaries come from different clocks

    An animation notification, effect timer, network event, and physics update can disagree at blends, low frame rates, pause, rollback, or seeking. Drive collision, boundary flashes, sound, result labels, and replay data from the same deterministic attack time, then assert the frames immediately before, at, and after both boundaries.

Tuning

Tune the attack

  • Represent the attack as ordered states

    Use startup → active → follow-through → recovery rather than a single attacking flag. Define legal transitions and cancellation ownership, and make state entry perform activation or cleanup exactly once. This keeps hitboxes from leaking when the boss is interrupted, dies, changes phase, or loses its target.

  • Separate duration from coverage

    A longer active interval and a larger hitbox solve different design problems. Tune duration, reach, width, movement, and rotation independently; then test early, middle, and late contact against the slowest supported response and several starting distances.

  • Choose a repeat-hit policy

    One active interval may contain many simulation ticks. Give each attack instance an ID and record targets already affected, or define an intentional repeat cadence. Do not let frame rate, overlapping hurtboxes, or hit-stop multiply damage accidentally.

  • Keep debug truth available

    Expose the current state, attack-instance ID, collision shape, contact set, and boundary timestamps in developer tools and replays. A designer should be able to pause on 1.59, 1.60, 2.17, and 2.18 seconds and explain every visual and collision result without guessing.

Arena

  • A rune seam that burns for one measured beat

    Kern draws force into a rectangular seam across the floor. The seam is broken and cool during startup, burns as one solid band only during the active interval, then fractures into a harmless afterimage while Kern's stone arm completes its swing. The material change explains the collision boundary without adding a separate HUD.

Player upgrades

  • Let Tavi act on both sides of the boundary

    Tavi needs no dash or invulnerability: ordinary movement clears the full 48-unit lane before activation. After the end flash, the same movement crosses the still-visible strike and reaches sword range during recovery. Faster verbs may increase margin, but they must not redefine when the hitbox exists.

Do not confuse this with

  • Damage interval, not preparation or downtime

    Wind-up is the readable preparation before release; attack lock freezes selected parameters; the active phase is only the interval in which the attack can apply its effect; recovery is the post-action period before the boss can act again. A visible motion may span all four, so animation length alone does not define any boundary.

Bosses that use this mechanic

Learn more

Found an inaccuracy or have a suggestion for this page? Propose an edit on GitHub