tdmcp

English

Português

English

Português

Appearance

Layer-1 generators reference ​

The prompt cookbook shows these generators in action — grouped by what you want to make. This page is the companion reference: one short "what it builds + when to reach for it" entry per generator, so you can pick the right altitude when two tools look similar (the GPU particle field versus the flock; the color grade versus the color wheels; kinetic text versus the crawl).

Layer-1 generators each build a whole, wired, previewable network in one call. Most expose live controls and default to a self-contained synthetic source so they preview with no camera, no audio device and no plugin. You rarely name them — the AI picks them from your prompt — but knowing the vocabulary helps you steer.

Generators that belong to a dedicated arc (pose/hands, show transport, front-of-house, MediaPipe) are listed below with a one-liner and a link to their guide rather than re-described here.

Generative & abstract ​

  • create_strange_attractor — Integrates a chosen ODE system (Lorenz / Aizawa / Halvorsen) into a glowing, ever-drawing polyline. Reach for it when you want deterministic, mathematically organic line-flow. The CPU-geometry sibling of create_growth_system (L-systems) and create_particle_flock (boids).
  • create_sdf_field — A programmable signed-distance-field raymarcher in a single GLSL TOP: a CSG tree of spheres / boxes / tori with smooth blending and live camera/step/color controls. Reach for it for clean procedural 3D forms with no geometry pipeline.
  • create_generative_art — The catch-all evolving visual: picks a technique (reaction-diffusion, noise landscape, attractor, voronoi, fractal, custom GLSL) and exposes a Speed knob. Reach for it when you want "something generative and moving" without committing to a specific algorithm.
  • create_jfa_voronoi — A Jump-Flooding Voronoi / stained-glass cell pattern in GLSL, with palette, seed-count, jitter and edge controls. Reach for it for crisp cellular tessellation looks.
  • create_reaction_diffusion — A Gray-Scott reaction-diffusion GPU sim with live F / K / Da / Db sliders and coral/spots/stripes/mitosis LUT presets. Reach for it for organic, growing, biological patterning.
  • create_simulation — Reaction-diffusion, slime, or fluid-style feedback trails with a Decay knob. Reach for it for drifting, advected, "alive" texture; step up to create_generative_art for more procedural techniques.
  • create_volumetric_field — A stacked-slice fake-volumetric noise field (smoke / nebula / ember / ice / toxic / mono) accumulated Beer-Lambert style. Reach for it for atmospheric depth without a true raymarcher.
  • create_voxel_stack — An isometric voxel-stack renderer driven by any TOP's luminance (Monument-Valley pastel ramps included). Reach for it for blocky, data-driven 3D relief.

Particles & 3D ​

  • create_gpu_particle_field — A high-count GPU point field (up to ~262k) simulated in feedback-TOP loops, optionally energised by audio or camera motion. Reach for it for dense particle drift well beyond the CPU emitter.
  • create_particle_flock — The boids variant of the GPU field: separation / alignment / cohesion with live knobs. Reach for it for flocking behaviour rather than curl-noise/gravity drift.
  • image_to_particles — Turns an image into a particle field where each particle's rest position and colour is its pixel; audio scatters and re-forms it — the "image dissolves into points on the drop" look.
  • create_depth_displacement — Pushes a subdivided plane into true 3D relief by a depth/luminance map. Reach for it for a 2.5D landscape that shifts with the camera (use create_depth_silhouette for a flat 2D mask instead).
  • create_depth_pop_field — A depth-driven GPU POP scatter field that samples a depth/mask TOP for displacement and emission; auto-spins a MediaPipe segmentation chain if you give it no depth source.
  • create_depth_silhouette — Extracts a white-on-black body silhouette/mask from a depth or video source, optionally colour-filled. Reach for it for camera-reactive installations and masks.
  • create_depth_from_2d — Wraps the community TDDepthAnything v2 TOX to turn any 2D image/video into a depth map (NVIDIA GPU only). Feed its output into the depth tools above. Requires the user-installed TOX.
  • create_gaussian_splat_scene — Drops the community TDGS TOX and loads a .ply/.splat Gaussian-Splat asset with camera binding. Requires the user-installed TOX.

Audio-reactive ​

  • create_audio_reactive — The all-in-one: an audio analysis chain plus a spectrum visual (bars / radial / particle / feedback / LED grid). Reach for it when you want a finished audio-driven visual in one call.
  • create_transient_reactive — Splits audio into normalized transient (percussion) and sustain (tonal floor) channels for bind_to_channel. Reach for it to drive different visuals from hits versus pads.
  • create_chroma_reactive (experimental) — A 12-channel pitch-class chroma vector for harmonic/key-aware binding.
  • create_energy_structure (experimental) — A self-calibrating song-structure detector emitting energy, state (breakdown/build/drop) and edge pulses. Reach for it to trigger scene changes off the arrangement, not just the beat.
  • create_midi_note_reactive — A MIDI-note → per-note velocity/trigger chain with bindable note0…noteN channels. Reach for it to drive visuals per key.
  • audio_fingerprint_to_visual — Samples a few seconds of audio, fingerprints it, and auto-picks and dispatches a matching generator. Reach for it as a one-shot "match the music" starter.

Camera & motion reactive ​

  • create_optical_flow — A cheap CPU motion-energy field (bright = motion) built from stock TOPs, a drop-in for displacement/particle chains. Not a true dense-flow solver — reach for it for fast, no-CUDA motion texture.
  • create_blob_reactive — Tracks the positions of multiple blobs/hands in a camera and exposes blob0_x, blob0_y, blob0_size, … for binding. The per-object counterpart to single-value motion reactivity.
  • create_vector_lines — A pulse-driven image-to-vector-lines system (Trace SOP) for clean line-art overlays. Intentionally non-realtime: you press Vectorize to update.

Body-pose reactivity lives in its own arc — see Body & pose tracking:

  • create_pose_reactive — Derives scalar reactive channels (hand height, arm openness, elbow angle, velocity) from a tracked pose for bind_to_channel.
  • create_pose_controlnet_driver — Renders a canonical OpenPose stick figure TOP from a pose CHOP, ready to send to a ControlNet / Stable-Diffusion node.

Video & camera, scopes ​

  • create_video_scopes — A broadcast monitor (waveform / RGB parade / vectorscope / histogram). See Front-of-house dashboard.
  • create_waveform — A time-domain audio oscilloscope. See Front-of-house dashboard.
  • create_histogram_scope — A standalone luminance (and per-channel RGB) histogram scope for any TOP, ready for preview or bind_to_channel.

Text & titles ​

  • create_text_overlay — Composites styled text over a visual (or on transparency) — a finished title/lyric/credits layer ready for setup_output.
  • create_kinetic_text — A single word/line that flashes, pulses or slides, the signature live-VJ lyric-flash. Reach for it for animated typography; bind the LFO to a beat CHOP to lock flashes to tempo.
  • create_text_crawl — Multi-line scrolling ticker / vertical credits roll / typewriter reveal. Reach for it for multi-line copy and continuous scrolling (use create_kinetic_text for a single animated string).
  • create_text_3d — Extruded 3D glyphs with optional spin — title cards and 3D text drops. Use create_kinetic_text instead for flat 2D animated text.

Signature effects & looks ​

  • create_slit_scan — The "time-as-space" slit-scan look: each output row samples a different past frame from a Cache ring buffer.
  • create_pixel_sort — Kim-Asendorf-style luminance-thresholded pixel-sort streaks, with luma/hue/saturation sort keys.
  • create_ascii_render — Turns any TOP into a character-grid ASCII render (phosphor-green CRT default).
  • create_dither — Retro ordered-Bayer / error-diffusion dithering to a 2/4/16 colour palette (Game-Boy duotone default).
  • create_chrome_blobs — A liquid-chrome / Y2K metaball generator with metal tints and animated specular highlights.
  • create_kaleidoscope — Folds a source into N mirrored wedges with live Segments / Rotation / Zoom / Center.

Colour & finishing ​

  • create_color_grade — A lift/gamma/gain + HSV + optional LUT finishing stage — the "make the final output look graded" tool.
  • create_color_wheels — Three-way (shadows/mids/highlights) colour wheels with per-channel lift/gamma/gain. Reach for it when create_color_grade is too coarse.
  • apply_post_processing — Chains several distinct effects (bloom, glitch, rgb_split, vignette, …) in series over a source. Reach for it to stack effects; use a dedicated single-effect tool when you want its own exposed controls.
  • enhance_build — Scores a build and asks the LLM for allowlisted tool calls to raise the weakest sub-scores (optionally auto-applying them).

Shaders & imports ​

  • import_isf_shader — Imports an ISF (.fs) shader as a GLSL TOP with auto-generated controls (raw source, file, or URL).
  • import_shadertoy — Builds a GLSL TOP from a Shadertoy URL / ID / pasted source, wiring iChannels and exposing Speed/Mouse.

Installations & studies ​

  • moodboard_to_system — Ingests 1–6 moodboard images and builds a matching generative system (palette + motion + generator pick) via the vision LLM or a deterministic grammar.
  • create_facade_mapping — A multi-projector architectural facade rig (per-projector crop / corner-pin / edge-blend), shipped as a calibration skeleton. See Output & mapping.
  • create_kinect_wall_harp — A projected wall instrument with laser-like strings, two-hand wall-touch tracking, an OSC Kinect input mode and an internal pluck synth. Reach for it when a depth camera and projector become the instrument, and follow the physical installations checklist before making live tracking claims.
  • create_test_pattern — A projector calibration/alignment source (grid, crosshair, colour bars, ramp, circle-grid) with optional per-projector ID overlay.

Performance & automation ​

These build engines and lanes rather than visuals — see Show timelines & setlists for the full transport arc:

  • create_autopilot — A beat-driven auto-VJ that, every N beats, randomizes a target's controls or cycles its stored cues.
  • create_automation_lane — Records a live parameter sweep into a ring buffer over N bars, then loops it back on a bar-phase clock.
  • create_chop_recorder — Captures a source CHOP over a window and plays the take back (persisted across reloads), ready for bind_to_channel.
  • compose_cue_list, create_setlist_runner, create_phrase_locked_cue_engine, create_safety_blackout_chain — covered in Show timelines & setlists.

Body, hands & MediaPipe ​

Covered in their own guides:

  • create_pose_tracking, create_pose_skeleton, create_body_reactive, create_hand_hologram — see Body & pose tracking.
  • setup_body_tracking and the face/hand/segmentation adapters — see MediaPipe adapters.

Output & mapping ​

  • setup_output — Routes a finished TOP to a window, NDI, Syphon/Spout, recording or Touch Out — usually the last step.
  • create_multi_output — Fans a master TOP across N projectors with optional edge-blend feathering. See Output & mapping.

See also ​