
godot-3d-lighting
Expert patterns for Godot 3D lighting including DirectionalLight3D shadow cascades, OmniLight3D attenuation, SpotLight3D projectors, VoxelGI vs SDFGI, and LightmapGI baking. Use when implementing realistic 3D lighting, shadow optimization, global illumination, or light probes. Trigger keywords: DirectionalLight3D, OmniLight3D, SpotLight3D, shadowenabled, directionalshadowmode, directionalshadowsplit, omnirange, omniattenuation, spotrange, spotangle, VoxelGI, SDFGI, LightmapGI, ReflectionProbe, Environment, WorldEnvironment.
- Installs
- 452
- Last 24h
- —
- Trending rank
- —
- Security
- Not audited
Install
Works with every agent the skills CLI supports.
npx skills add https://github.com/thedivergentai/game-dev-agentic-skills --skill godot-3d-lighting- 01
Install
Run the command in your project. The CLI asks which agents to install for.
- 02
Describe the task
Ask your agent for something this skill covers. It loads godot-3d-lighting automatically when relevant.
- 03
Review the result
The agent follows the SKILL.md steps, scripts and references bundled with the skill.
SKILL.md
The instructions your agent reads when the skill is triggered.
3D Lighting
Expert guidance for realistic 3D lighting with shadows and global illumination.
NEVER Do
- NEVER use VoxelGI without setting a proper extents — Unbound VoxelGI tanks performance. Always set
sizeto tightly fit your scene. - NEVER enable shadows on every light — Each shadow-casting light is expensive. Use shadows sparingly: 1-2 DirectionalLights, ~3-5 OmniLights max.
- NEVER forget directional_shadow_mode — Default is ORTHOGONAL. For large outdoor scenes, use PARALLEL_4_SPLITS for better shadow quality at distance.
- NEVER use LightmapGI for fully dynamic scenes — Lightmaps are baked. Moving geometry won't receive updated lighting. Use VoxelGI or SDFGI instead.
- NEVER set omni_range too large — Light attenuation is quadratic. A range of 500 affects 785,000 sq units. Keep range as small as visually acceptable.
- NEVER hide a Light node using the Visible property to exclude it from a Lightmap bake — Hiding a light has no effect on the baker. You must change the light's Bake Mode to Disabled.
- NEVER use VoxelGI with paper-thin walls — VoxelGI evaluates lighting using a 3D grid. Thin walls (less than one voxel thick) will cause severe light leaking. Seal your geometry or place hidden thick MeshInstance3D blocks around the exterior.
- NEVER leave shadow bias at default for cascades — Default bias often causes Peter Panning or light leaking at split transitions. Tune bias per-light based on your scene's scale.
- NEVER bake LightmapGI without a Denoiser — Godot's baked lightmaps are noisy by default. Use OIDN or JNLM (in Project Settings) for professional results.
- NEVER use real-time SDFGI on Mobile/Compatibility renderers — It is a Forward+ exclusive feature. Use fake GI bounce lights for lower-end platforms.
- NEVER use 'Update Continuity' in ReflectionProbes for performance — Keep ReflectionProbes on 'Update Once' and trigger manual updates only when necessary.
- NEVER overflow the clustered shadow atlas — WHY: Forward+ packs Omni/Spot shadows into a shared atlas. Too many shadowed positional lights → atlas thrash, flickering, or silent quality collapse. Cap shadowed Omni (~3–5) and Spot (~2–4); use
light_lod_optimizer.gd/ distance fade before raising atlas size. - NEVER stack dozens of overlapping shadowed Omni/Spot in one cluster cell — WHY: Clustered light iteration cost scales with overlapping lights per tile. Prefer non-shadowed fills, tighter
omni_range/spot_range, anddistance_fade_*so far lights leave the cluster.
Available Scripts
MANDATORY: Read the appropriate script before implementing the corresponding pattern.
Golden path (pick one — Do NOT Load peers)
| Scene | Load | Do NOT Load |
|---|---|---|
| Outdoor Forward+ | shadow_cascade_tuner.gd + sdfgi_probe_manager.gd (+ day_night_cycle.gd if time-of-day) | fake_gi_bounce.gd, indoor VoxelGI bake unless hybrid; environment_blender.gd / fog unless atmosphere task |
| Indoor sealed geometry | MANDATORY VoxelGI via light_probe_manager.gd + lighting_manager.gd + ReflectionProbes | SDFGI on tiny interiors; Mobile fake GI unless targeting Mobile; sky/fog scripts unless zone blend |
| Mobile / Compatibility | MANDATORY fake_gi_bounce.gd + light budgets in light_lod_optimizer.gd | Real-time SDFGI (Forward+ only); HDR sky/volumetric body recipes |
| Lightmap bake / hybrid static | MANDATORY lightmap_bake_helper.gd (+ Shadowmasking for outdoor) | Runtime SDFGI as bake substitute; disabling lights via Visible |
| Sky / Environment blend only | environment_blender.gd | Cascade/GI managers when only tonemap/ambient/sky changes |
| Volumetric fog / shafts | volumetric_fx.gd (+ volumetric_fog_zones.gd for localized density) | Pure light-budget / shadow-atlas tuning tasks |
| Pure light budget / shadow LOD | light_lod_optimizer.gd, shadow_bias_tuner.gd | Do NOT Load environment_blender.gd, volumetric_fx.gd, day-night unless required |
day_night_cycle.gd
Dynamic sun position and color based on time-of-day. Handles DirectionalLight3D rotation, color temperature, and intensity curves. Use for outdoor day/night systems.
light_probe_manager.gd
VoxelGI and SDFGI management for global illumination setup.
lighting_manager.gd
Dynamic light pooling and LOD. Manages light culling and shadow toggling based on camera distance. Use for performance optimization with many lights.
volumetric_fx.gd
Volumetric fog and god ray configuration. Runtime fog density/color adjustments and light shaft setup. Use for atmospheric effects.
shadow_cascade_tuner.gd
Expert logic for adjusting DirectionalLight3D shadow split distances dynamically based on sun angle and camera tilt.
lightmap_bake_helper.gd
Advanced LightmapGI configuration pattern using Shadowmasking mode for hybrid static/dynamic shadowing.
sdfgi_probe_manager.gd
Dynamic quality scaler for real-time Global Illumination (SDFGI). Adjusts cell size and occlusion for performance/quality trade-offs.
volumetric_fog_zones.gd
Smoothly transitioning localized fog density for cave entrances or forest clearings using Tweens and Area3D triggers.
fake_gi_bounce.gd
Efficient 'Mobile-GI' pattern. Simulates light bouncing off the floor using non-shadowed directional fill lights.
environment_blender.gd
Architectural pattern for transitioning WorldEnvironment parameters (Sky, Ambient, Tonemap) during gameplay.
shadow_bias_tuner.gd
Optimization script for correcting 'Peter Panning' and 'Shadow Acne' on high-fidelity directional lights.
light_lod_optimizer.gd
Distance-based shadow and visibility culling for OmniLight3D nodes in dense environments.
reflection_probe_manager.gd
Performance-aware ReflectionProbe handling using manual 'Update Once' triggers for large environmental changes.
spotlight_projector_setup.gd
High-detail lighting using Projector textures to fake complex shadow patterns (grates, glass ripples).
light_volume_trigger.gd
Area3D-driven camera Environment blend for cave/desert transitions (duplicate env resource; tween exposure/ambient).
lighting_quality_manager.gd
Runtime RenderingServer profile: half-res SDFGI, shadow atlas size, volumetric fog density.
DirectionalLight3D (Sun/Moon)
Shadow Cascades
# For outdoor scenes with camera moving from near to far
extends DirectionalLight3D
func _ready() -> void:
shadow_enabled = true
directional_shadow_mode = SHADOW_PARALLEL_4_SPLITS
# Split distances (in meters from camera)
directional_shadow_split_1 = 10.0 # First cascade: 0-10m
directional_shadow_split_2 = 50.0 # Second: 10-50m
directional_shadow_split_3 = 200.0 # Third: 50-200m
# Fourth cascade: 200m - max shadow distance
directional_shadow_max_distance = 500.0
# Quality vs performance
directional_shadow_blend_splits = true # Smooth transitions
Day/Night Cycle
MANDATORY day_night_cycle.gd — do not re-inline sun energy/color recipes here.
OmniLight3D (Point Light)
Keep omni_range tight; prefer quadratic attenuation. Flicker/campfire loops belong in scene scripts — not this body. Shadowed Omni count is a hard budget (see NEVER).
SpotLight3D (Flashlight/Headlights)
MANDATORY spotlight_projector_setup.gd for range/angle/projector cookies and camera-follow flashlights. Do not paste Spot setup here.
Global Illumination (script-first)
Use the golden-path table above. Body decision only:
| Path | Script | When |
|---|---|---|
| Indoor / sealed | MANDATORY light_probe_manager.gd (+ lighting_manager.gd) | Tight VoxelGI extents per room; never paper-thin walls |
| Outdoor Forward+ | MANDATORY sdfgi_probe_manager.gd | Real-time GI; Do NOT Load on Mobile/Compatibility |
| Static / Mobile bake | MANDATORY lightmap_bake_helper.gd | LightmapGI + Shadowmasking; bake mode ≠ Visible hide |
| No GI budget | MANDATORY fake_gi_bounce.gd | Fill lights only |
Do not re-inline VoxelGI/SDFGI/Lightmap property setup here.
Environment & Sky
Sky/ambient/tonemap transitions → MANDATORY environment_blender.gd. Volumetric fog / shafts → MANDATORY volumetric_fx.gd (+ volumetric_fog_zones.gd for caves/forests).
Do NOT Load these for pure light-budget / shadow-atlas / cascade tuning — stay on light_lod_optimizer.gd / shadow_cascade_tuner.gd.
ReflectionProbe
For localized reflections (mirrors, shiny floors):
# reflection_probe.gd
extends ReflectionProbe
func _ready() -> void:
# Capture area
size = Vector3(10, 5, 10)
# Quality
resolution = ReflectionProbe.RESOLUTION_512
# Update mode
update_mode = ReflectionProbe.UPDATE_ONCE # Bake once
# or UPDATE_ALWAYS for dynamic reflections (expensive)
Performance Optimization
Light Budgets
# Recommended limits:
# - DirectionalLight3D with shadows: 1-2
# - OmniLight3D with shadows: 3-5
# - SpotLight3D with shadows: 2-4
# - OmniLight3D without shadows: 20-30
# - SpotLight3D without shadows: 15-20
# Disable shadows on minor lights
@onready var candle_lights: Array = [$Candle1, $Candle2, $Candle3]
func _ready() -> void:
for light in candle_lights:
light.shadow_enabled = false # Save performance
Per-Light Shadow Distance
# Disable shadows for distant lights
extends OmniLight3D
@export var shadow_max_distance := 50.0
func _process(delta: float) -> void:
var camera := get_viewport().get_camera_3d()
if camera:
var dist := global_position.distance_to(camera.global_position)
shadow_enabled = (dist < shadow_max_distance)
Edge Cases
Shadows Through Floors
# Problem: Thin floors let shadows through
# Solution: Increase shadow bias
extends DirectionalLight3D
func _ready() -> void:
shadow_enabled = true
shadow_bias = 0.1 # Increase if shadows bleed through
shadow_normal_bias = 2.0
Light Leaking in Indoor Scenes
# Problem: VoxelGI light bleeds through walls
# Solution: Place VoxelGI nodes per-room, don't overlap
# Also: Ensure walls have proper thickness (not paper-thin)
Expert Techniques & Optimizations
1. Shadowmasking for Large Outdoor Scenes
Rendering real-time shadows for distant objects is too expensive. Use Shadowmasking by setting a DirectionalLight3D to the Dynamic bake mode while baking a LightmapGI. This bakes distant shadows into a texture while allowing dynamic objects to cast real-time shadows up close, preventing "double shadowing" artifacts.
2. Fake Global Illumination
If you cannot afford GI at all (e.g., strict mobile constraints), fake it! Duplicate your main DirectionalLight3D, rotate it 180 degrees (pointing up from the ground), turn Shadows OFF, set Specular to 0.0, and reduce Energy to 10-40%. This cheaply simulates bounced floor lighting.
3. Simulating PCSS (Contact-Hardening Shadows)
Godot's OmniLight3D scaling can simulate Percentage-Closer Soft Shadows (blurrier shadows further from the caster).
extends OmniLight3D
func _ready() -> void:
# Simulates area lights and Percentage-Closer Soft Shadows (PCSS).
# Note: High performance cost. Keep the number of lights with light_size > 0.0 low.
light_size = 0.5
shadow_enabled = true
# Distance fade culls the light and shadow completely when out of range,
# preventing the clustered renderer from choking on too many overlapping PCSS lights.
distance_fade_enabled = true
distance_fade_begin = 20.0
distance_fade_length = 5.0
Expert Pattern: Light-Volume-Trigger
Smoothly transition between lighting environments (e.g., entering a dark cave from a bright desert) using Area3D triggers and Tween-driven Camera3D overrides.
class_name LightVolumeTrigger extends Area3D
@export var interior_environment: Environment
@export var transition_duration: float = 2.0
func _ready() -> void:
body_entered.connect(_on_body_entered)
body_exited.connect(_on_body_exited)
func _on_body_entered(body: Node3D) -> void:
if body.is_in_group("player"):
var camera := get_viewport().get_camera_3d()
# Duplicate to avoid modifying the original resource
if not camera.environment:
camera.environment = interior_environment.duplicate()
var tween := create_tween().set_parallel(true)
# Interpolate key properties for visual adaptation
tween.tween_property(camera.environment, "tonemap_exposure", interior_environment.tonemap_exposure, transition_duration)
tween.tween_property(camera.environment, "ambient_light_energy", interior_environment.ambient_light_energy, transition_duration)
func _on_body_exited(body: Node3D) -> void:
if not body.is_in_group("player"):
return
var camera := get_viewport().get_camera_3d()
if camera == null or camera.environment == null:
return
var world_env := get_tree().get_first_node_in_group("world_environment") as WorldEnvironment
var target := world_env.environment if world_env and world_env.environment else Environment.new()
var tween := create_tween().set_parallel(true)
tween.tween_property(camera.environment, "tonemap_exposure", target.tonemap_exposure, transition_duration)
tween.tween_property(camera.environment, "ambient_light_energy", target.ambient_light_energy, transition_duration)
tween.chain().tween_callback(func():
# Return to WorldEnvironment ownership when blend completes
camera.environment = null
)
[!TIP] Place a
ReflectionProbeinside the interior withinterior = true. Godot will automatically blend this with the exterior environment as the player transitions.
Expert Pattern: Interior-Mapping (Fake-Rooms)
Fake interior depth on window quads is a custom spatial-shader specialty — implement in a project shader or route to godot-shaders-basics. Do not keep incomplete ray-box stubs in this skill body.
Expert Pattern: Lighting-Quality-Settings
Manage complex lighting features (Shadows, SDFGI, Fog) at runtime using the RenderingServer API for direct engine control.
class_name LightingQualityManager extends Node
func apply_low_quality_profile(env_rid: RID) -> void:
# 1. SDFGI Optimization
# Huge performance gain: Render GI buffers at half resolution
RenderingServer.gi_set_use_half_resolution(true)
RenderingServer.environment_set_sdfgi_ray_count(RenderingServer.ENV_SDFGI_RAY_COUNT_4)
RenderingServer.environment_set_sdfgi_frames_to_converge(RenderingServer.ENV_SDFGI_CONVERGE_IN_30_FRAMES)
# 2. Shadow Optimization
# Reduce global directional shadow atlas
RenderingServer.directional_shadow_atlas_set_size(2048, true)
RenderingServer.directional_soft_shadow_filter_set_quality(RenderingServer.SHADOW_QUALITY_SOFT_VERY_LOW)
# Reduce positional (Omni/Spot) shadows for current viewport
get_viewport().positional_shadow_atlas_size = 1024
# 3. Volumetric Fog
# Disable or heavily reduce fog detail
RenderingServer.environment_set_volumetric_fog(env_rid, false, 0.01, Color.WHITE, Color.BLACK, 0.0, 0.2, 64.0, 2.0, 1.0, true, 0.9, 0.0, 1.0)
Deep recipes (on demand)
| Topic | Reference / script |
|---|---|
| VoxelGI / SDFGI / LightmapGI setup | gi-and-bake-recipes.md |
| Sky / fog / HDR environment | environment-and-fog.md |
Reference
Progressive disclosure: open Official Documentation links only when researching a specific API; load Related Skills when routing to a peer domain — do not preload the whole lattice.
Official Documentation
- Introduction to lights and shadows — Directional/Omni/Spot roles, shadow atlas budgets, cascades, projectors, and bias tradeoffs.
- Introduction to global illumination — Decision matrix for LightmapGI vs VoxelGI vs SDFGI vs probes before committing bake or runtime cost.
- Using SDFGI — Forward+-only real-time GI cascades, occlusion, and quality knobs used by outdoor/open scenes.
- Using VoxelGI — Bounded indoor GI, subdivision, leak causes (thin walls), and per-room bake workflow.
- Using LightmapGI — Static bake modes, denoiser, shadowmasks, and when baked lighting cannot serve dynamic geometry.
- Reflection probes — Localized specular capture, Update Once vs Always, and interior blending with environments.
- Faking global illumination — Fill-light / bounce approximations for Mobile and Compatibility when SDFGI is unavailable.
- Environment and post-processing — WorldEnvironment sky, ambient, tonemap, and camera overrides for light-volume transitions.
- Volumetric fog and fog volumes — Global volumetric fog plus FogVolume density zones for shafts and localized atmosphere.
- Physical light and camera units — Lux/EV-style energy when matching real-world day/night and exposure chains.
- Optimizing 3D performance — Shadow/GI/light-count budgets after local LOD and cascade tuning still miss frame time.
- AreaLight3D — Rectangular soft area lights for panels/screens instead of emissive-only fakes in Forward+.
Related Skills
Prerequisites
- godot-project-foundations — Renderer choice (Forward+/Mobile/Compatibility), shadow atlas, and GI project settings gate which lighting features exist.
- godot-3d-materials — Albedo/roughness/metallic/emission and
gi_modedecide how surfaces receive bounced and baked light. - godot-gdscript-mastery — Typed node scripts, Tweens, and Resource ownership patterns used by managers and environment blends.
Complements
- godot-shaders-basics — Custom spatial/fog/sky shaders and interior-mapping tricks that extend Environment lighting without more real lights.
- godot-camera-systems — Camera
environmentoverrides and exposure pairing for cave/interior light-volume transitions. - godot-3d-world-building — Scene scale, sealed interiors, and GridMap/static layout that make VoxelGI/LightmapGI bake cleanly.
- godot-particles — Dust/smoke/embers read correctly only when fog density and light energy are co-authored.
- godot-tweening — Smooth Environment, fog, and light-energy transitions instead of hard cuts at Area3D boundaries.
- godot-physics-3d — Area3D body enter/exit triggers that drive light-volume and fog-zone blends.
- godot-adapt-desktop-to-mobile — Strip SDFGI/heavy shadows and swap to fake bounce GI when targeting Mobile/Compatibility.
Downstream / consumers
- godot-performance-optimization — Escalate here when cascade splits, probe counts, or volumetric fog still dominate the profiler.
- godot-monte-carlo-balancer — Simulate visibility/readability under day-night or darkness budgets when lighting changes combat, stealth, or horror difficulty.
- godot-genre-horror — Consumes sparse lights, fog, and ReflectionProbe interiors for tension-first atmospheres.
- godot-debugging-profiling — Use debugger/monitor views to prove shadow atlas and GI cost before cutting features blindly.
Master
- godot-master — Library router and mirrored module entry; open when discovering which Domain Skill owns a cross-cutting lighting concern.
FAQ
What does the godot-3d-lighting skill do?
Expert patterns for Godot 3D lighting including DirectionalLight3D shadow cascades, OmniLight3D attenuation, SpotLight3D projectors, VoxelGI vs SDFGI, and LightmapGI baking. Use when implementing realistic 3D lighting, shadow optimization, global illumination, or light probes. Trigger keywords: DirectionalLight3D, OmniLight3D, SpotLight3D, shadowenabled, directionalshadowmode, directionalshadowsplit, omnirange, omniattenuation, spotrange, spotangle, VoxelGI, SDFGI, LightmapGI, ReflectionProbe, Environment, WorldEnvironment.
How do I install godot-3d-lighting?
Run npx skills add https://github.com/thedivergentai/game-dev-agentic-skills --skill godot-3d-lighting in your project. The skills CLI detects the agents you use and copies the skill into the right folder. For Claude Code you can also place the skill folder at ~/.claude/skills/godot-3d-lighting/ (personal) or .claude/skills/godot-3d-lighting/ (project).
Which AI agents can use godot-3d-lighting?
Any agent that supports the open Agent Skills format (a folder with a SKILL.md file), including Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI, Windsurf, Cline, OpenCode and more.
Is godot-3d-lighting safe to install?
No partner security audit is available yet. Skills can run commands and scripts, so read the SKILL.md and any bundled files before installing.
Who maintains godot-3d-lighting?
It is published by thedivergentai in thedivergentai/game-dev-agentic-skills. It has 452 recorded installs.
More from thedivergentai/game-dev-agentic-skills
godot-master
thedivergentai/game-dev-agentic-skills
Consolidated expert library for professional Godot 4.7+ game and application development. Orchestrates 92 Domain Skills through architectural workflows, anti-pattern catalogs, performance budgets, and Server API patterns. Use when: (1) starting a new Godot project, (2) designing game or app architecture, (3) building entity/component systems, (4) debugging performance or physics issues, (5) choosing between 2D/3D approaches, (6) implementing multiplayer, (7) optimizing draw calls or script time, (8) porting between platforms, (9) migrating from 4.6 to 4.7. Primary entry point for ALL Godot development tasks. Keywords: Godot 4.7, AreaLight3D, HDR, Asset Store, godot-master.
godot-shaders-basics
thedivergentai/game-dev-agentic-skills
Expert Godot shader patterns for batch-safe hitflash, alpha-scissor foliage/dissolve, screenspace postFX, depth reconstruction, triplanar, and instance uniforms — not first-shader tutorials. Trigger on draw-call batching breaks, discard vs depth-prepass, foliage shadows, post-process quads, or world-position FX. Keywords: instance uniform, ALPHASCISSOR, hintscreentexture, hintdepthtexture, global uniform, sampler2DArray, canvasitem, spatial, post-processing.
godot-2d-animation
thedivergentai/game-dev-agentic-skills
Expert patterns for 2D animation in Godot using AnimatedSprite2D and skeletal cutout rigs. Use when implementing sprite frame animations, procedural animation (squash/stretch), cutout bone hierarchies, or frame-perfect timing systems. Trigger keywords: AnimatedSprite2D, SpriteFrames, animationfinished, animationlooped, framechanged, frameprogress, setframeandprogress, cutout animation, skeletal 2D, Bone2D, procedural animation, animation state machine, advance(0).
godot-ui-theming
thedivergentai/game-dev-agentic-skills
Expert blueprint for UI themes using Theme resources, StyleBoxes, custom fonts, and theme overrides for consistent visual styling. Covers StyleBoxFlat/Texture, theme inheritance, dynamic theme switching, and font variations. Use when implementing consistent UI styling OR supporting multiple themes. Keywords Theme, StyleBox, StyleBoxFlat, addthemeoverride, font, theme inheritance, dark mode.
godot-gdscript-mastery
thedivergentai/game-dev-agentic-skills
Expert GDScript landmine guidance: static typing opcodes, signal-up/call-down, %UniqueName/@onready lifecycle, Callable bind/unbind, await sequences, typed collections, and safe Dictionary iteration. Use for code review, refactoring hot paths, or project standards. Trigger keywords: statictyping, signalarchitecture, uniquenodes, @onready, classname, signalupcalldown, Callable.bind, typedcollections, awaitsequence.
godot-particles
thedivergentai/game-dev-agentic-skills
Expert blueprint for GPU particle systems (explosions, magic effects, weather, trails) using GPUParticles2D/3D, ParticleProcessMaterial, gradients, sub-emitters, and custom shaders. Use when creating VFX, environmental effects, or visual feedback. Keywords GPUParticles2D, ParticleProcessMaterial, emissionshape, colorramp, subemitter, oneshot.
Popular media & creative skills
hyperframes
heygen-com/hyperframes
Read first to make, edit, animate or render any video with HyperFrames (HTML rendered to video), including launch films, explainers, music videos, charts, maps, vertical clips, title cards, captions and overlays. Holds the render contract, the work loop and what each kind of video needs.
media-use
heygen-com/hyperframes
Get or make any media for a HyperFrames video - music, sound effects, images, icons, logos, voiceover, avatar and generated video, transcripts, cutouts, colour grades and looks - with HeyGen or free local tools, and mix placed audio. Also for footage that looks dark, flat or dull, or should be stylized or hidden.
hyperframes-registry
heygen-com/hyperframes
About 450 ready-made HyperFrames blocks and components installed as editable source - transitions, overlays, titles, kinetic type, captions, charts, map overlays, 3D, camera moves. Use it before hand-building any named visual (CRT scanlines, glitch, film grain, a chart); also for hyperframes add and catalog.
hyperframes-creative
heygen-com/hyperframes
Non-animation creative direction for HyperFrames videos. Use for design spec (frame.md / design.md) handling, palettes, typography, narration, beat planning, audio-reactive visuals, composition patterns, and brand / style decisions. For atomic motion patterns and scene blueprints, use hyperframes-animation.
ai-video-generation
101-skills/superpowers
Generate AI videos with Google Veo, Seedance 2.0, HappyHorse, Wan, Grok and 40+ models via inference.sh CLI. Models: Veo 3.1, Veo 3, Seedance 2.0, HappyHorse 1.0, Wan 2.5, Grok Imagine Video, OmniHuman, Fabric, HunyuanVideo. Capabilities: text-to-video, image-to-video, reference-to-video, video editing, lipsync, avatar animation, video upscaling, foley sound. Use for: social media videos, marketing content, explainer videos, product demos, AI avatars. Triggers: video generation, ai video, text to video, image to video, veo, animate image, video from image, ai animation, video generator, generate video, t2v, i2v, ai video maker, create video with ai, runway alternative, pika alternative, sora alternative, kling alternative, seedance, happyhorse
ai-image-generation
101-skills/superpowers
Generate AI images with GPT-Image-2, FLUX, Gemini, Grok, Seedream, Reve and 50+ models via inference.sh CLI. Models: GPT-Image-2, FLUX Dev LoRA, FLUX.2 Klein LoRA, Gemini 3 Pro Image, Grok Imagine, Seedream 4.5, Reve, ImagineArt. Capabilities: text-to-image, image-to-image, inpainting, LoRA, image editing, upscaling, text rendering. Use for: AI art, product mockups, concept art, social media graphics, marketing visuals, illustrations. Triggers: flux, image generation, ai image, text to image, stable diffusion, generate image, ai art, midjourney alternative, dall-e alternative, text2img, t2i, image generator, ai picture, create image with ai, generative ai, ai illustration, grok image, gemini image, gpt image, openai image, chatgpt image