Viewport logo

Viewport

Community
kevinpbuckley
viewport

Control the Unreal Editor level viewport — camera type/position, view mode, FOV, exposure, layout, and rendering settings (ViewportService). Use when the user asks to move the editor camera, change the view mode (Lit/Unlit/Wireframe), set FOV/exposure, switch viewport layout, or frame the level for a screenshot.

Overview

Publisherkevinpbuckley
RepositoryVibeUE
Skill nameviewport
Stars
735
Forks
161
Bundled files
1
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 1 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by kevinpbuckley on GitHub. Read the source before you install it.

Installation

Install the Viewport AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/kevinpbuckley/VibeUE/tree/master/Content/Skills/viewport
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/kevinpbuckley/VibeUE.git /tmp/VibeUE
mkdir -p .claude/skills
cp -r /tmp/VibeUE/Content/Skills/viewport .claude/skills/viewport
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Viewport in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Viewport on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Viewport is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Viewport Control Skill

Methods

MethodDescription
get_viewport_info()Get full viewport state (type, camera, FOV, exposure, layout)
set_viewport_type(type)Switch perspective/orthographic views
get_viewport_type()Get current view type as string
set_view_mode(mode)Switch rendering mode (lit, wireframe, unlit, etc.)
get_view_mode()Get current rendering mode
set_fov(degrees)Set field of view (5-170, default 90)
get_fov()Get current FOV
set_near_clip_plane(distance)Set near clipping plane (-1 = engine default)
set_far_clip_plane(distance)Set far clipping plane (0 = infinity)
set_exposure(fixed, ev100)Set fixed/auto exposure with EV100 value
set_exposure_game_settings()Reset exposure to auto (game settings)
set_game_view(enable)Toggle Game View (hides editor icons)
set_allow_cinematic_control(allow)Allow Sequencer to control viewport camera
set_realtime(enable)Toggle realtime rendering
set_camera_speed(speed)Set camera movement speed (1-8)
set_viewport_layout(name)Switch viewport layout (single, quad, etc.)
get_viewport_layout()Get current layout name
capture_scene(location, rotation, width, height, out_png, ortho_width=0, fov=90, manual_ev100=0)Synchronous SceneCapture2D → PNG that works while the editor is backgrounded (see below)

🔀 Camera get/set moved to the engine. ViewportService no longer has set_camera_location / set_camera_rotation. Reading and positioning the viewport camera is now Unreal 5.8's native EditorAppToolset (GetCameraTransform / SetCameraTransform, plus FocusOnActors to frame actors) — reach it with call_tool (run describe_toolset on EditorAppToolset for exact action names/params). ViewportService still owns view type, view mode, FOV, clip planes, exposure, game view, layout, realtime, and camera speed.

Critical Rules

⚠️ A black capture usually means no lighting, not a broken mesh

The most common cause of an all-black (or blown-out white) capture is the level, not the camera and not the asset you just built. Two things to check before you start debugging geometry:

  1. A level made with new_level() has no lighting at all — no sun, no sky, no atmosphere. Use new_level_from_template(path, "/Engine/Maps/Templates/Template_Default") instead (see the level-actors skill). If you must light an empty level by hand you need a DirectionalLight that genuinely points down — unreal.Rotator(roll=0, pitch=-48, yaw=125), and note the argument order is (roll, pitch, yaw), so a positionally-passed pitch aims the sun at the sky and the scene stays black — plus a SkyLight with real_time_capture, and a SkyAtmosphere.
  2. Auto-exposure lies about colour. With adaptation on, a dark material renders near-white and a bright one renders grey, so you cannot judge a material from a capture. Add an unbound PostProcessVolume with auto_exposure_method = AEM_MANUAL; then auto_exposure_bias is a brightness stop — higher is brighter, and the usable window is narrow (0 was pitch black and 13.5 blown out in one scene; 7.5 was correct). Tune it by capturing, not by reasoning.

Keep the volume out of frame (move it far below the set) — an unbound volume still applies globally but its bounds box draws in the editor viewport.

⚠️ Valid View Types for set_viewport_type

Only these exact strings are accepted (case-insensitive):

StringView
"perspective"Perspective (3D)
"top"Top-down (XY)
"bottom"Bottom-up (-XY)
"left"Left side (-XZ)
"right"Right side (XZ)
"front"Front (-YZ)
"back"Back (YZ)
python
# ✅ CORRECT
unreal.ViewportService.set_viewport_type("perspective")
unreal.ViewportService.set_viewport_type("top")

# ❌ WRONG — these are not valid strings
unreal.ViewportService.set_viewport_type("ortho")
unreal.ViewportService.set_viewport_type("iso")

⚠️ Valid View Modes for set_view_mode

StringRendering Mode
"lit"Default fully lit
"unlit"No lighting
"wireframe"Wireframe overlay
"detaillighting"Detail lighting
"lightingonly"Lighting only (no textures)
"lightcomplexity"Light complexity heatmap
"shadercomplexity"Shader complexity heatmap
"pathtracing"Path tracing
"clay"Clay rendering

⚠️ Valid Layout Names for set_viewport_layout

NameLayout
"OnePane"Single viewport (default)
"TwoPanesHoriz"Two side-by-side
"TwoPanesVert"Two stacked
"ThreePanesLeft"Large left + 2 right
"ThreePanesRight"Large right + 2 left
"ThreePanesTop"Large top + 2 bottom
"ThreePanesBottom"Large bottom + 2 top
"FourPanesLeft"Large left + 3
"FourPanesRight"Large right + 3
"FourPanesTop"Large top + 3
"FourPanesBottom"Large bottom + 3
"FourPanes2x2"Quad view (2x2 grid)
"Quad"Alias for FourPanes2x2

🚨 FOV Only Works in Perspective Mode

Setting FOV has no effect in orthographic views. Always switch to perspective first:

python
unreal.ViewportService.set_viewport_type("perspective")
unreal.ViewportService.set_fov(75.0)

🚨 Realtime Mode vs On-Demand

When set_realtime(False), the viewport only repaints on interaction. All ViewportService methods force a redraw after changes, so this is transparent to Python callers — but be aware users won't see continuous animation/particles until realtime is re-enabled.

⚠️ Multi-pane read-back caveat: get_viewport_info().is_realtime reflects the active pane. In a multi-pane layout (e.g. FourPanes2x2) where the active pane isn't the primary, set_realtime(True) can succeed yet is_realtime reads back False. Verify realtime in OnePane layout, or don't rely on the read-back to gate logic when in a split layout. (Other fields like view type / FOV / camera read back correctly across layouts.)

📸 Capture that works while backgrounded

CaptureViewport / CaptureEditorImage (and HighResShot) return a stale frame when the MCP-driven editor is minimised or backgrounded — the level viewport does not pump frames. The reliable path is ViewportService.capture_scene(...): it spawns a transient ASceneCapture2D, renders one frame synchronously with CaptureScene(), reads the pixels back and writes the PNG in-call. No viewport pumping, no editor_invalidate_viewports() + sleep dance.

python
import unreal

# Perspective grab of a spot in the level (works even when the editor window is hidden)
res = unreal.ViewportService.capture_scene(
    unreal.Vector(1200, -800, 900),          # camera location
    unreal.Rotator(pitch=-20, yaw=45, roll=0),  # kwargs — Rotator positional order is (roll,pitch,yaw)
    1280, 720,                                # width, height (px)
    "C:/temp/shot.png",                       # absolute path, or relative → Saved/VibeUE/Captures
    manual_ev100=-6.0)                        # 0 = auto (black when hidden); negative EV = brighter fixed exposure — see below
if res.b_success:
    print("wrote", res.output_path, res.file_size_bytes, "bytes")
else:
    print("failed:", res.error_message)

Baked-in facts (measured — these are why the hand-rolled recipe existed):

  • Alpha. The capture source is SCS_FINAL_COLOR_LDR, which yields alpha 255. SCS_BASE_COLOR writes alpha 0, producing a PNG that renders as a blank white page in most viewers even though the file is a full-size capture. capture_scene forces the exported alpha to 255 regardless, so its PNGs are always opaque. (Check the file size, not just the preview — a truly empty PNG is a few KB.)
  • Format. The render target is RTF_RGBA8. The engine default (RTF_RGBA16f, a float format) writes non-PNG bytes; capture_scene never uses it.
  • Exposure. Leave manual_ev100=0 (default) to keep the engine's automatic exposure — fine for a foreground / PIE window, but black when the editor is hidden (no converged eye adaptation). A backgrounded editor needs a fixed exposure: pass manual_ev100 != 0 and capture_scene switches to a manual exposure decoupled from the physical camera, where the value is the exposure TARGET in EV100 — exactly like a camera's metered EV: higher EV100 is DARKER (it assumes a brighter scene and stops down), lower/negative is brighter. Measured on a daylit scene from a backgrounded editor (mean RGB luminance): 0 (auto) → 16.4, +4 → 4.3, -4 → 43.8, -6 → 63.4, -8 → 86.0. A daylit backgrounded scene reads well around -6 to -8; try -4 first for bright scenes. Tune by capturing.

    Note: the engine reads its underlying AutoExposureBias as a pow(2,bias) exposure compensation (so its docs say "positive brightens"), but in this decoupled-Manual SceneCapture path the observed, repeatable behaviour is the opposite — higher = darker — so treat manual_ev100 as a camera EV target. Do not carry over any old "bias ~10–14 to brighten" advice; that was from a build that left the physical camera coupled, and here a large positive value is near-black.

  • Cleanup. The transient capture actor and its render target are destroyed inside the call — no stray actors are left in the level.

Minimap / top-down map recipe (orthographic, north-up):

python
import unreal
# Ortho capture centred over the map, looking straight down, north-up.
# ortho_width = the world-space span you want to cover (e.g. the landscape size in uu).
map_size = 500000.0  # uu across
res = unreal.ViewportService.capture_scene(
    unreal.Vector(0, 0, 100000),                    # high above centre; Z only needs to clear geometry
    unreal.Rotator(pitch=-90, yaw=-90, roll=0),     # pitch=-90 looks down; yaw=-90 makes +X point up = north-up
    2048, 2048,
    "minimap.png",                                  # → Saved/VibeUE/Captures/minimap.png
    ortho_width=map_size,
    manual_ev100=-6.0)   # fixed exposure for a hidden editor; LOWER (more negative) = brighter
  • ortho_width > 0 selects orthographic projection (and fov is ignored); ortho_width == 0 (default) is perspective using fov.
  • The rotation (pitch=-90, yaw=-90, roll=0) yields a north-up image (world +X points to the top). This is a convention of the caller's rotation, not something the API forces — pass a different yaw to rotate the map.
  • For flat map labels, a TextRenderActor at rotation (roll=0, pitch=90, yaw=90) reads correctly in this north-up view.

Workflows

Inspect Current Viewport State

python
import unreal
info = unreal.ViewportService.get_viewport_info()
view_mode = unreal.ViewportService.get_view_mode()
print(f"Type: {info.viewport_type}")
print(f"View Mode: {view_mode}")
print(f"Location: {info.location}")
print(f"Rotation: {info.rotation}")
print(f"FOV: {info.fov}")
print(f"Layout: {info.layout}")
print(f"Realtime: {info.is_realtime}")
print(f"Game View: {info.is_game_view}")

Cycle Through Orthographic Views

python
import unreal
for view in ["top", "front", "right", "perspective"]:
    unreal.ViewportService.set_viewport_type(view)

Switch to Quad View and Back

python
import unreal
# Switch to quad (2x2) layout
unreal.ViewportService.set_viewport_layout("Quad")

# Switch back to single pane
unreal.ViewportService.set_viewport_layout("OnePane")

Set Up Architecture Review Camera

python
import unreal
# Perspective with narrow FOV for minimal distortion
unreal.ViewportService.set_viewport_type("perspective")
unreal.ViewportService.set_fov(60.0)
unreal.ViewportService.set_game_view(True)
unreal.ViewportService.set_exposure(True, 1.0)  # Fixed exposure
unreal.ViewportService.set_realtime(True)

Wireframe Debugging

python
import unreal
# Switch to wireframe to inspect mesh topology
unreal.ViewportService.set_view_mode("wireframe")

# Return to normal lit view
unreal.ViewportService.set_view_mode("lit")

Position Camera at Specific Location

Camera placement now lives in the engine EditorAppToolset (SetCameraTransform), called via call_tool. From Python you can still drive it through the editor subsystem:

python
import unreal
# Move camera to a bird's-eye view (location, rotation)
# ⚠️ Rotator positional order is (Roll, Pitch, Yaw) — always use kwargs
unreal.get_editor_subsystem(unreal.UnrealEditorSubsystem).set_level_viewport_camera_info(
    unreal.Vector(0, 0, 5000), unreal.Rotator(pitch=-90))

To frame specific actors instead of guessing coordinates, prefer the engine EditorAppToolset.FocusOnActors action (or load the level-actors skill for ActorService.get_actor_view_camera, which auto-computes a framing position from bounds).

Configure Clipping Planes

python
import unreal
# Tighten near clip for close-up work
unreal.ViewportService.set_near_clip_plane(1.0)

# Set far clip to avoid rendering distant objects
unreal.ViewportService.set_far_clip_plane(50000.0)

# Reset to defaults
unreal.ViewportService.set_near_clip_plane(-1)  # Engine default
unreal.ViewportService.set_far_clip_plane(0)     # Infinity

Exposure Control

python
import unreal
# Fix exposure for consistent lighting review
unreal.ViewportService.set_exposure(True, 1.0)

# Adjust to bright/dark scene
unreal.ViewportService.set_exposure(True, -2.0)  # Darker
unreal.ViewportService.set_exposure(True, 4.0)   # Brighter

# Return to auto exposure (game settings)
unreal.ViewportService.set_exposure_game_settings()

Sample scripts (run via execute_python_code)

  • scripts/set_camera.txt — position the editor camera and set the view mode.

Additional gotchas

  • CaptureViewport through call_tool needs BOTH optional params present ({"captureTransform":{}, "annotations":{}}); captureTransform is ignored (it always frames the world origin), and the ~1 MB base64 result must be decoded to a file before it can be viewed.
  • Outside PIE the MCP-driven viewport does not pump frames, so captures are stale — static actors render but dynamic FX are absent and HighResShot never fires. Call PerformanceService.set_background_throttling(False) first; asset thumbnails still render fine while backgrounded.
  • CaptureEditorImage returns the previous frame unless you call editor_invalidate_viewports() and wait briefly first.
  • Use capture_scene for a real off-screen render instead of hand-building a SceneCapture2D; a backgrounded editor has no converged eye adaptation, so fix the exposure (AEM_MANUAL, bias ~10-14) or the shot comes out black.
  • Never A/B a capture against one taken earlier in the session — streaming, LOD and lighting drift between moments; after a material change, recompile and discard one warm-up capture or the thumbnail shows the default checker.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Viewport AI skill do?

Control the Unreal Editor level viewport — camera type/position, view mode, FOV, exposure, layout, and rendering settings (ViewportService). Use when the user asks to move the editor camera, change the view mode (Lit/Unlit/Wireframe), set FOV/exposure, switch viewport layout, or frame the level for a screenshot.

Why use Viewport on TypingMind?

Because you install it once and use it with any model. Viewport is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Viewport in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/kevinpbuckley/VibeUE/tree/master/Content/Skills/viewport. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Viewport?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Viewport?

As many as you like. As long as a model supports skills, you can use Viewport with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Viewport AI skill free?

Yes. It is published on GitHub by kevinpbuckley under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇