PyVista Cheat Sheet

1Setup & load0.48

import pyvista as pv mesh = pv.read("model.stl")★
One reader for everything — STL, OBJ, PLY, VTK/VTU/VTP, gLTF and more, via a pluggable reader registry. Returns a mesh object.
from pyvista import examples mesh = examples.download_bunny() # built-in sample data★
The examples module has dozens of downloadable datasets — great for learning and reproducing docs.
pip install "pyvista[all]" # + jupyter, io, imageio extras
Base pyvista covers desktop plotting; the [all] extra adds Jupyter (trame), extra I/O, and movie export.

2Create meshesprimitives & from arrays

pv.Sphere() · pv.Cube() · pv.Cylinder() · pv.Plane() · pv.Cone()★
Parametric primitives return PolyData — handy for tests, glyphs, and scene props.
cloud = pv.PolyData(points) # (N, 3) numpy -> point cloud surf = pv.PolyData(points, faces) # + connectivity -> surface★
Build from NumPy. points is an (N,3) array; faces is a VTK-style connectivity array. pv.wrap(arr) wraps NumPy/VTK objects.
grid = pv.ImageData(dimensions=(10, 10, 10)) # uniform grid pv.StructuredGrid(x, y, z) · pv.UnstructuredGrid(...)
Gridded types for volumes/FE data. Note: ImageData is the current name for the old UniformGrid.

3Data arrayspoint_data vs cell_data

mesh.point_data["temp"] = temps # per-vertex (len == n_points) mesh.cell_data["region"] = ids # per-cell (len == n_cells)★
Attach NumPy arrays to points or cells. Length must match n_points/n_cells respectively.
mesh["temp"] = temps # shortcut (infers point/cell by length) mesh.set_active_scalars("temp")★
The mesh["name"] shortcut assigns by length; the active scalars are what plots/filters use by default.
mesh.points · mesh.n_points · mesh.n_cells · mesh.bounds · mesh.center
Geometry introspection. mesh.points is a live (N,3) NumPy view — edit it to move vertices.

4Plotter & add_meshrender a scene

mesh.plot(scalars="temp", cmap="coolwarm", show_edges=True)★
Quick one-liner for a single mesh — opens an interactive window.
pl = pv.Plotter() pl.add_mesh(mesh, scalars="temp", cmap="viridis", opacity=0.7) pl.show()★
The full API. Add as many meshes/actors as you like, then show(). add_mesh returns the actor for later tweaks.
pl.add_mesh(mesh, style="wireframe", color="red", line_width=2) pl.add_points(cloud, render_points_as_spheres=True, point_size=8)
style: "surface" / "wireframe" / "points". Use a flat color= instead of scalars= for a solid look.

5Filtersreturn NEW meshes

clipped = mesh.clip("x") # cut by a plane sl = mesh.slice("z") # cross-section th = mesh.threshold((0.2, 0.8)) # keep cells in range★
Filters return a new mesh — assign the result. Also clip_box, slice_orthogonal(), clip_surface.
iso = mesh.contour([0.5]) # isosurfaces of active scalar warped = mesh.warp_by_scalar() # displace by scalar★
contour extracts level sets; warp_by_scalar/warp_by_vector deform geometry by a field.
arrows = mesh.glyph(orient="vectors", scale="mag", factor=0.1) lines = mesh.streamlines(vectors="velocity")
Vector-field viz: glyph places oriented arrows; streamlines traces flow. Also decimate, smooth, extract_surface, compute_normals.
result = mesh.clip("x").slice("y").elevation() # chainable
Because each filter returns a mesh, they chain fluently.

6Scalars & colorbarscolor mapping

pl.add_mesh(mesh, scalars="temp", cmap="jet", clim=(0, 100), below_color="grey", above_color="red")★
Map a field to color. cmap takes any Matplotlib/colorcet map; clim fixes the range; out-of-range gets below/above colors.
pl.add_scalar_bar(title="Temp (°C)", n_labels=5)★
Customize the colorbar; suppress it with add_mesh(..., show_scalar_bar=False).
pl.add_mesh(mesh, scalars="class", categories=True, annotations={0: "air", 1: "tissue"})
For discrete labels, categories=True + annotations gives a categorical colorbar.

7Views, widgets, volumericher scenes

pl = pv.Plotter(shape=(1, 2)) pl.subplot(0, 1); pl.add_mesh(other) pl.link_views()★
Side-by-side subplots with a shape grid; link_views() syncs their cameras.
pl.add_mesh_clip_plane(mesh) # drag a cutting plane pl.add_mesh_slice(mesh) · pl.add_mesh_threshold(mesh)★
Interactive widgets — adjust clip/slice/threshold live with on-screen handles.
pl.add_volume(vol, cmap="bone", opacity="sigmoid")★
Volume rendering of 3-D scalar fields (ImageData) — medical/CT data. opacity takes a named transfer function or an array.
pl.add_axes(); pl.show_grid(); pl.camera_position = "xy"
Orientation axes, bounds grid, and preset camera views ("xy", "iso"…).

8Export & Jupyterimages, HTML, notebooks

pl.show(screenshot="scene.png") · pl.screenshot("out.png")★
Save a raster image of the view. Set the window size via Plotter(window_size=(1920,1080)).
pl.export_html("scene.html")★
Standalone interactive HTML (VTK.js) — pan/zoom/rotate in any browser, no Python needed.
pv.set_jupyter_backend("trame") # interactive in notebooks★
Jupyter: "trame" = interactive widget; "static" = image; "html" = embedded scene. Needs the jupyter extras installed.
pl.open_gif("spin.gif") for ang in range(360): pl.camera.azimuth = ang; pl.write_frame()
Animations to GIF/MP4 (open_movie) — write a frame per step.

!Common gotchasread before shipping

filters return NEW meshes
mesh.clip("x") doesn't modify mesh — it returns a clipped copy. Assign the result or it's lost.
show() blocks & resets
pl.show() renders then closes the render window; add all meshes before calling it, and build a fresh Plotter to draw again.
scalar length must match
A point array needs length n_points, a cell array n_cells. The mesh["x"] shortcut guesses by length — be explicit with point_data/cell_data when ambiguous.
headless servers need off_screen + Xvfb
No display = show() fails. Use pv.OFF_SCREEN = True / Plotter(off_screen=True) and a virtual framebuffer (Xvfb) or EGL. pv.start_xvfb() helps on Linux.
Jupyter needs a backend
No interactive widget until pv.set_jupyter_backend("trame") and pyvista[jupyter]/trame are installed.
UniformGrid -> ImageData
pv.UniformGrid was renamed to pv.ImageData. Old tutorials use the former; use ImageData.
huge meshes = slow interaction
Rendering is GPU-bound. decimate/extract_surface, use point clouds, or export static images/HTML for very large data.

PyVista cheat sheet