shapeviz 0.1.0

Lightweight 3D mesh and point cloud viewer that opens an interactive browser tab. No Open3D, no VTK, no heavy dependencies.

shapeviz

A lightweight 3D mesh and point cloud viewer that opens an interactive browser tab.

No Open3D. No VTK. No 500 MB install that breaks every time you change Python versions. Just pip install shapeviz, point it at a file, call view(), and a beautiful interactive 3D viewer opens in your browser — powered by Three.js loaded straight from a CDN.

import shapeviz

shapeviz.view("mesh.ply")
shapeviz.view("cloud.xyz", mode="pointcloud", point_size=2)
shapeviz.compare("before.ply", "after.ply")   # side by side

---

Why shapeviz?

If you work with 3D geometry in Python — robotics, point-cloud ML, NeRF, mesh processing, game assets — your options for just looking at the thing are heavy. Open3D and VTK are powerful but large, native, and notoriously fragile across Python/OS versions.

shapeviz is the opposite:

	shapeviz	Open3D
Install size	a few KB	~500 MB
Required dependencies	zero	many native libs
Rendering	Three.js in your browser	native OpenGL window
Works over SSH / in notebooks	✅ (browser / inline IFrame)	✗ awkward
Breaks on Python upgrade	basically never	often

numpy and scipy are optional — install them only if you want faster mesh processing. Everything works without them.

---

Installation

pip install shapeviz

Optional extras:

pip install "shapeviz[numpy]"   # faster normal computation / heatmaps
pip install "shapeviz[full]"    # numpy + scipy
pip install "shapeviz[dev]"     # contributor tooling (pytest, ruff, build…)

Requires Python 3.9+. Works on Windows, macOS, and Linux.

---

Supported formats

Extension	Type	ASCII	Binary
.ply	mesh / point cloud	✅	✅
.obj	mesh (with vertex colors)	✅	—
.stl	mesh	✅	✅
.xyz, .xyzrgb, .pts	point cloud (XYZ / XYZRGB / XYZ+normals)	✅	—
.pcd	point cloud	✅	✅

The loader auto-detects meshes vs. point clouds and reads vertex colors and normals when present.

---

Python API

view(source, **options)

Open a single mesh or point cloud.

import shapeviz

# From a file path...
shapeviz.view("bunny.ply")

# ...or from an in-memory object you built yourself.
from shapeviz import Mesh
m = Mesh(vertices=[[0,0,0],[1,0,0],[0,1,0]], faces=[[0,1,2]])
shapeviz.view(m)

Common options (all keyword-only):

Option	Values	Description
mode	"solid", "wireframe", "pointcloud", "normals"	initial render mode
color_mode	"solid", "vertex", "normals", "heatmap"	initial coloring
point_size	float	point size for clouds
color	hex string, e.g. "#6cb6ff"	solid color
background	hex string	background color
show_axes	bool	show XYZ axes on start
show_grid	bool	show the ground grid
compute_normals	bool	compute per-vertex normals for meshes that lack them
open_browser	bool	open the system browser (default True)
inline	bool	force/disable inline Jupyter rendering (default: auto)
return_html	bool	return the raw HTML string instead of viewing

shapeviz.view("cloud.xyz", mode="pointcloud", point_size=3, color_mode="heatmap")
shapeviz.view("part.stl", color="#ff8866", compute_normals=True)

compare(*sources, titles=None, **options)

Show two or more geometries side by side with synchronised cameras — rotate one, they all rotate. Perfect for before/after, ground-truth vs. prediction, raw vs. filtered.

shapeviz.compare("before.ply", "after.ply")
shapeviz.compare(raw, filtered, meshed, titles=["raw", "filtered", "meshed"])

save_html(source, path, **options)

Render to a standalone .html file you can email, embed, or open later. (The only external thing it needs is the Three.js CDN at view time.)

shapeviz.save_html("scan.pcd", "report.html", color_mode="heatmap", point_size=2)

load(path)

Just parse a file into geometry without rendering.

g = shapeviz.load("mesh.ply")
print(g)                       # <Mesh 'mesh.ply' vertices=35947 faces=69451 ...>
print(g.num_vertices, g.bounds())

---

Interactive viewer controls

Once the viewer opens, a floating panel gives you:

• Render mode — solid · wireframe · point cloud · normals overlay
• Color mode — solid color · vertex colors · normals-as-color · heatmap (by Z)
• Solid color picker and point size slider
• Toggles for grid, axes, and a light/dark background
• Sync cameras (in compare mode)
• Reset view and Screenshot (saves a PNG)

Mouse: drag to rotate · right-drag to pan · scroll to zoom.

---

Command line

shapeviz installs a shapeviz command:

shapeviz view mesh.ply
shapeviz view cloud.xyz --mode pointcloud --point-size 3 --color-mode heatmap
shapeviz compare before.ply after.ply
shapeviz save mesh.ply out.html --color "#ff8866"
shapeviz info cloud.pcd          # print stats without rendering

shapeviz view/compare start a tiny local HTTP server and open your browser; press Ctrl+C to stop. Add --no-browser to start the server without opening a tab. Run shapeviz --help or shapeviz <command> --help for all options.

---

Jupyter notebooks

Inside a notebook, view() and compare() render inline automatically via an IFrame — no server, no new tab:

import shapeviz
shapeviz.view("mesh.ply", height=500)

Use inline=False to force a browser tab instead, or inline=True to force inline rendering outside a notebook.

---

How it works

1. Parse the file in pure Python into a Mesh or PointCloud.
2. Serialise vertices/faces/colors/normals into a compact JSON payload.
3. Embed that payload into a self-contained HTML document that pulls Three.js from a CDN and builds a BufferGeometry in the browser.
4. Serve it from Python's built-in http.server (or inline it as a data-URI IFrame in Jupyter, or write it to disk with save_html).

No native rendering, no GPU drivers to fight with — your browser already has a great WebGL renderer.

---

Development

git clone https://github.com/amanmukati09/shapeviz.git
cd shapeviz
python -m venv .venv
.venv\Scripts\activate           # Windows
# source .venv/bin/activate      # macOS/Linux
pip install -e ".[dev]"
pytest
ruff check src tests

The test suite generates tiny sample files for every format on the fly and runs both with and without numpy, so it verifies the pure-Python fallbacks too.

---

License

MIT © Tushar Mukati. See LICENSE.