Text to Motion

The core generation card. The prompt text itself is authored on the Prompt Timeline (card 06); this card holds the generation parameters, the output target, and the Generate Motion button. The header shows the current Duration (frames · seconds) and reminds you that a single prompt maxes out at 10 s.

#General

#Advanced (toggle to enable)

Mirrors the fuller parameter set. The toggle in the header shows/hides the body; its Reset restores the parameters without collapsing the panel.

#Output Target — the generation mode

The Mode segment is the single most important choice on this card: it decides where the generated animation is written and how it coexists with the rest of your scene. All three modes generate from the same prompts and parameters — they differ only in the destination and its side effects. Pick the mode first, then the mode-specific options appear directly beneath it.

#New Take

Creates a brand-new MotionBuilder take and writes the generated motion into it, leaving all your existing takes untouched. This is the safest default while you iterate: every generate produces a clean, self-contained clip you can compare or throw away.

• Auto-name from first prompt (on by default) — the take is named from the first N characters of the earliest timeline prompt (N is set by Auto take-name length in Settings), so takes are self-describing. Regenerating the same prompt adds a _v001 / _v002… version suffix instead of overwriting, giving you an automatic version history.
• Turn auto-naming off to type an exact Take name of your own.
• Num samples > 1 fans out into that many separate takes in one run, so you can audition several variations of the same prompt side by side.

#Existing Take

Applies the generated motion into the take that is currently active in MotionBuilder, shown as a read-only Current take display (Refresh re-reads the active take). Use this to iterate on a specific take, to overwrite a previous attempt in place, or to layer generated motion onto animation you're already working in — rather than accumulating new takes.

• Because everything merges into one existing take, Num samples is forced to 1 here (multiple samples have nowhere distinct to land) — the field is greyed out.
• Combine with Generate on base layer only (Layer & Rig) to control whether keys land on the base layer or the active animation layer of that take.

#Story

Writes each prompt block as its own FBX clip on a MotionBuilder Story track (one track per block, exported to your Story Export Path from Settings). This is the non-linear, editorial mode: instead of one baked take, you get discrete clips you can re-order, trim, blend, and re-time on the Story timeline — ideal for sequencing a performance out of several actions and revising individual beats without regenerating the whole thing.

• Regenerating a block writes a new version of that clip and keeps the track in place, so the Story edit you built survives across regenerations.
• Passthrough — gaps between clips reveal the underlying take's animation instead of freezing, so the character keeps moving between generated beats.
• Replace current Clip — overwrite the newest version of a clip in place rather than adding another version, keeping the folder tidy while you dial one beat in.
• Baking to the Control Rig is a no-op in Story mode (the motion lives in a passthrough clip, not on skeleton keys), so that option is greyed out here.

Which mode when — Reach for New Take while exploring (clean, versioned, disposable), Existing Take to refine or overwrite one specific take, and Story to assemble and non-linearly edit a multi-action performance from separate clips.

#Placement

Re-seats the generated (origin-canonicalized) motion to the character's scene position at the prompt's start frame.

• Use current position (on by default) — start from the skeleton's scene XZ and facing instead of the origin. It owns horizontal placement and yaw only, never height.
• Preserve height (off by default) — starts the motion at the height the rig is actually standing at rather than the model's own standing height. It is sent to the server as a start anchor on the hips (a height, not a local offset), so nothing is added to the result afterwards — what you see in the viewport is what the pin means. Independent of Use current position: either can be on without the other. A Full Body constraint covering the start frame takes precedence, since it already carries a height of its own.

Scope of Preserve height — It applies to the main Generate Motion run. It is not sent by Generate Selected Prompt or by the per-gap fan-out (when your prompt blocks have gaps between them, each contiguous group is generated as its own run) — the Console says so when you generate with the option on. On a single-block regeneration, leaving Keep start pose checked achieves the same thing, because the pinned boundary pose already carries the hip height.

#Ground

A movable ground-plane object whose world-Y is folded into the vertical offset, so motion stands at ground + hip height.

• Create Ground — spawn or find the plane.
• Ground height (m) — the plane is the source of truth (the field spans −1000 … 1000 m and stays disabled until a plane exists); Refresh pulls its live Y into the field, and editing the field moves the plane.
• Hide / Show — toggle plane visibility.
• Apply ground offset (on by default) — fold the plane's height into the generated motion so the character stands on the plane. Applies in all modes. The related Correct ground offset switch lives in Settings → Extra options.

#Layer & Rig

Generate Motion builds MMCP segments from the active character's timeline blocks and sends the request. A live progress bar and status label show the run; the card is locked while generating. Shortcut: Ctrl+G over the timeline.