tn-venv vs venv, virtualenv, and uv

An honest feature comparison. Where another tool is better, this page says so.

tn-venv

stdlib venv

virtualenv

uv venv

Zero runtime dependencies

✅

✅

❌ (distlib, filelock, platformdirs)

✅ (static binary)

Written in

Python

Python

Python

Rust

pip installed by default

✅

✅

✅

❌

setuptools/wheel seeding

✅ optional

❌ (3.12+)

✅

❌

Pin pip at creation (--pip X.Y)

✅

❌

✅ (bundled wheel version)

❌

Offline seeding

✅ bundled + --extra-search-dir

bundled only

✅ app-data wheels

✅

--with PKG / -r reqs.txt at creation

✅

❌

❌

❌

Shells with activation scripts

7 (bash, batch, ps1, fish, csh, nu, activate_this)

3–4

5+

4

activate_this.py

✅

❌

✅

❌

Interpreter discovery

PATH, PEP 514, py, uv, pyenv

current interpreter only

PATH, registry

PATH, managed downloads

Can download missing Pythons

❌

❌

❌

✅

Config files + env vars

✅ pyproject.toml, ini, TN_VENV_*

❌

✅ ini + env

✅ env + config

--dry-run

✅

❌

❌

❌

Inter-process creation lock

✅

❌

✅ (app-data)

❌

Speed (cold create, Windows)

~2–4 s

~2–5 s

~0.5–1 s

~0.05 s

Guarantees identical output to python -m venv

near-identical by design

✅ by definition

❌

❌

When to choose what

  • python -m venv — you are on a machine where installing anything is impossible, and you do not need setuptools, extra shells, or discovery. venv is always there.

  • virtualenv — you need its app-data seed caching (faster repeated creation), its plugin ecosystem, or support for Python < 3.11 hosts.

  • uv venv — raw creation speed is the bottleneck (CI creating thousands of environments), or you want uv to download the interpreter too.

  • tn-venv — you want one memorable command with rigorous defaults, every shell’s activation script, per-project config in pyproject.toml, package installation at creation time, and a tool you can read end-to-end because it is pure, dependency-free Python.

Compatibility notes

Environments produced by tn-venv follow PEP 405 and are byte-comparable in structure to python -m venv output: same directory layout, same pyvenv.cfg keys (plus a few documented extras, see pyvenv.cfg reference), same binary placement on Windows (venvlauncher.exe redirectors on Python ≥ 3.11). Anything that consumes a standard virtual environment — IDEs, pip, tox, build frontends — treats a tn-venv environment exactly like a venv one.