QuickySitter’s core engine imposes no sitter-slot limit of its own, so the seat count you can wire in the AVpos notecard is bounded mainly by SL’s per-prim script-count cap. The product family does add caps, though: [QS]hudproxy hard-caps concurrent HUD-driven sitters at 6, and the [QS]faces, [QS]sequence and [QS]select plugins are sized for a small per-furniture cap (around 7 pre-handshake). So on a piece using the HUD or those plugins, plan around those limits rather than SL’s script cap. The setup is a direct extension of the Couples Setup procedure.
Working alone? The Creator Edition’s Animesh Adjust Dummies fill the empty seats of a group pose while you set it up: one posable dummy per seat, no extra avatars needed.
Adding more sitters
For N sitter slots:
[QS]boot(still just one instance)[QS]sitA,[QS]sitA 1,[QS]sitA 2, …,[QS]sitA <N-1>[QS]sitB,[QS]sitB 1,[QS]sitB 2, …,[QS]sitB <N-1>(matching count)
The suffixes start at ` 1 (the unnumbered script is slot 0) and must be **contiguous**: sitA counts its siblings by probing [QS]sitA 1, [QS]sitA 2`, … until the first gap, so a missing number silently truncates the slot count.
Each slot has its own sit-target offset, its own POSE/SYNC lines in the AVpos notecard, and its own row in qs:cfg:<ch> / qs:sitter:<ch> LSD keys. (ANIM is a separate directive, the [QS]faces face-animation command, not a per-slot pose line.)
SYNC across N slots
A SYNC pose with the same <menu_name> in N slots animates all N sitters in lockstep. Each slot picks its own per-slot animation file:
SITTER 0
SYNC Group-Hug|hug_left
SITTER 1
SYNC Group-Hug|hug_center
SITTER 2
SYNC Group-Hug|hug_right
SITTER <n> opens the section for sitter slot <n>; all subsequent POSE/SYNC/BUTTON lines belong to that slot until the next SITTER. See AVpos Reference for the full grammar, and the upstream multi-sitter docs for tutorial-style examples.
Re-Sync with N sitters
LinkMsg 90271 broadcasts to all sitA slots simultaneously via LINK_SET. Each slot re-phases its main animation in the same Sim frame; resulting drift is bounded to ~50 ms regardless of how many sitters are seated. See Re-Sync Protocol.
Memory considerations
At 1000+ poses across N slots, QuickySitter’s LSD-backed menu storage becomes a noticeable improvement over stock, because sitB stays slim while stock would push past the 64 KB Mono cap. The old in-RAM MENU_LIST was retired (0.9954); menu state is now built from paged LSD reads of qs:p:<ch>:<i>, windowed by the qs:nm sidecar. See LSD Storage.
The other scaling concern is [QS]offset’s RAM tier (200-entry LRU cap). With N users × N slots × M poses of personal offsets, persistent LSD storage (QSO:*) is the durable place; RAM is the volatile fallback. See Personal Pose Offsets.
Menu navigation
With many poses and many sitters, the dialog menu pagination becomes important. The [<<] / [>>] buttons page through entries when a section overflows; menu structure (top-level TOMENU buttons → submenus) helps group poses by scene.
See Submenus for navigation design.
Common gotchas
[QS]select. Optional and presence-gated, since sitB has a built-in seat picker, so multi-sitter menu routing works without it. When[QS]selectis present (sitB reads theqs:alive:selectflag, with a[AV]selectinventory probe as fallback), it provides the dedicated seat-select picker.- Seat prims. SL’s sit-target clamp is ±300 m per axis, so offsets are not the constraint on long furniture. Still, give each seat its own prim and pin it via the prim description (
<set>-<slot>): right-click Sit docks on the clicked prim, so separate seat prims are what make “sit where you click” work on a banquet table with 8 sitters. See SitTargets.
See also
- Couples Setup: the 2-sitter case.
- Animation Sequences: multi-step sequences across multiple sitters.
- Re-Sync Protocol: N-sitter sync.
- SitTargets: per-slot offsets.
