Character FX

Apply a character-driven effect (transformation, power, body modification, face FX, or subject-bound aura) to the subject of a connected AI video node.

Overview

The Character FX parameter node defines an effect that happens TO or AROUND the subject of an AI-generated video clip — the character transforms into something (werewolf, cyborg, vampire), demonstrates a fantastical power (fire-breathe, levitation, telekinesis), grows a body part (wings, horns, tentacles), shifts their face (oni mask, glowing eyes, X-ray), or is surrounded by an aura (paparazzi flashes, money rain, saint glow). The effect is described as natural-language prompt text and injected into the consumer video node’s prompt.

The distinguishing feature from other parameter pickers: when a character/face/object/location ref is wired to the node’s target input handle, every occurrence of "the subject" in the prompt is globally substituted with the ref’s display name. So [Character: "Aria Voss"] → target + werewolf becomes "Aria Voss transforms into a werewolf — fur sprouts across the skin..." instead of "the subject transforms...".

Configuration

Field	Type	Default	Description
Character FX	multi-select	`"auto"`	Catalog entry id, or array of 1-2 ids for compound effects (e.g., `["werewolf","fire-breathe"]`).
Position	select	`"auto"`	Where in the clip the effect occurs: `auto` / `start` / `middle` / `end` / `full`.
Duration	select	`"auto"`	How long the effect lasts: `auto` / `instant` / `short` (~1s) / `medium` (~2s) / `long` (~3s).
Intensity	select	`"auto"`	Energy/character of the effect: `auto` / `subtle` / `natural` / `dynamic` / `crazy`.
Pre Text	text	empty	Free-form text prepended to the composed hint.
Post Text	text	empty	Free-form text appended to the composed hint.

All four enum fields default to auto, which contributes no prompt text.

Catalog (57 entries across 5 categories)

Category	Examples	Theme
Transformation (14)	werewolf, vampire, cyborg, ghost-form, statue-stone, liquid-metal, animalization, gorilla-form, mystification, gas-form, diamond-skin, agent-reveal	Subject becomes a different entity
Power & Ability (12)	fire-breathe, ice-breathe, air/water/earth-bending, lightning-hands, levitation, telekinesis, invisibility, hero-flight, super-speed, soul-departure	Subject demonstrates a fantastical ability
Body Modification (9)	wings-grow, horns-grow, tail-emerge, tentacles-emerge, extra-eyes, head-explode (PG-13), head-off, spiders-from-mouth, skin-surge	Parts of the subject change, emerge, or leave
Face & Expression (8)	horror-face, oni-mask, glowing-eyes, floral-eyes, bloom-mouth, x-ray, agent-snap (sunglasses), visor-x	Face contorts, mask materializes, or eyes change
Aura & Ambient (14)	paparazzi, money-rain, color-rain, saint-glow, fire-aura, frost-aura, shadow-aura, electricity-aura, sparkles-around, fairies-around, objects-orbit, petals-around, glow-trace, tattoo-animation	Environmental FX bound to the subject

Two defaults round out the catalog: auto and none.

Multi-pick

Up to 2 effects can be selected and compounded. The composer runs subject substitution per id, then joins with ", and ". Examples:

["werewolf", "fire-breathe"] with target "Aria" — Aria transforms AND breathes fire.
["wings-grow", "glowing-eyes"] with target "Marcus" — Marcus grows wings AND his eyes ignite.

Target ref name substitution

The single input handle (target) accepts any of:

Character ref (data.characterName → name)
Face ref (data.faceName → name)
Object ref (data.objectName → name)
Location ref (data.locationName → name)

The fallback chain is: characterName ?? faceName ?? objectName ?? locationName ?? undefined. If target is unwired, every "the subject" in the prompt stays as "the subject" and the AI model fills in based on whatever subject is in frame.

Worked example with werewolf, position=middle, duration=long, intensity=dynamic, [Character: "Aria Voss"] → target:

“Aria Voss transforms into a werewolf — fur sprouts across the skin, fangs and claws extend, the snout elongates, eyes glow yellow, body re-shapes with visible muscle and bone movement under the skin, clothing tears, the effect occurs in the middle of the clip, manifesting over approximately 3 seconds, with dynamic energy and assertive flourish”

Inputs & Outputs

Inputs:

in — optional upstream parameter input.
target — optional, accepts a character / face / object / location ref node for subject name substitution.

Outputs:

out — composed prompt-hint clause, consumed by downstream AI video nodes via their cinematography handle.

Supported Providers

No direct provider call. Compatibility with downstream AI video providers:

Provider	Compatibility
VEO 3.1	Best for transformations + powers; coherent multi-frame morphing
Kling 2.1 / Kling Master	Strong for body-mod + aura effects
Hailuo Standard / 2.3	Good for face/expression + auras; transformations degrade
MiniMax	Limited transformation quality; auras work well
Seedance / Bytedance Lite	Hit-or-miss on transformation; auras reliable

Some effects (gas-form, liquid-metal, X-ray) are temporally complex and degrade noticeably on cheaper models. Prefer VEO 3.1 or Kling for those.

Best Practices

Wire a character ref to target for explicit subject naming — produces more grounded, character-consistent prompts.
Multi-pick (up to 2) for compound effects: transformation + power (werewolf + fire-breathe), body-mod + face (wings + glowing eyes).
For 3+ compound effects, wire two Character FX nodes in parallel into the consumer’s cinematography handle.
Pair with Camera Motion for choreography: e.g., Hero Flight effect + crash-zoom camera motion.
Set intensity: crazy for music-video moments where exaggeration is the goal; subtle for restrained narrative use.

Common Use Cases

Cinematic transformation arcs in AI-generated reels (werewolf, vampire, ghost-form).
Action sequences with elemental powers (fire-breathe, lightning-hands, water-bending).
Music video aura effects (money-rain, paparazzi, fire-aura, sparkles).
Horror/supernatural moments (spiders-from-mouth, extra-eyes, oni-mask).
Sci-fi reveals (cyborg, agent-reveal, X-ray, visor-x).

Tips

The composer rewrites "the subject" globally — apostrophe-s constructions like "the subject's body" become "Aria's body" naturally.
The auto and none entries have empty prompt hints — they contribute nothing and are silently filtered from multi-pick selections.
The Character FX node is automatically suppressed for still-image consumers (Edit Image, Image-to-Image, Generate Image) — character transformations are inherently temporal.
Each catalog entry includes a lucide-react icon in the picker grid (🐺 werewolf, 🔥 fire-breathe, 🎭 oni-mask, 👁️ extra-eyes, etc.) for quick visual identification.

Distinction from `action-fx`

Character FX — the SUBJECT does/becomes X: “the subject transforms into a werewolf”, “the subject breathes fire”.
Action FX — environmental events: “an earthquake hits the scene”, “lightning strikes the building”.

When in doubt: if the effect requires a character as the focal point, use Character FX. If it would make sense in an empty room, use Action FX. Both can be wired in parallel into the same consumer.