Python virtualenv module import fails despite successful pip installation

Building an automated video generation system using Python with PyTorch, WhisperX, MoviePy, and custom TTS pipelines, where the system combines text-to-speech with mechanical keyboard sound effects synchronized to video output.

Issue

The system failed with ModuleNotFoundError: No module named 'librosa' despite successful installation via uv pip install. The error occurred when running python main.py in a virtual environment where uv pip list confirmed librosa 0.11.0 was installed. The issue manifested as an import failure during execution while the package existed in the virtual environment’s site-packages.

Culprit

A Python version mismatch between the system Python (3.13.7) and the virtual environment (3.12.12) caused module resolution failures. The virtual environment was activated but:

1. The which python command pointed to /usr/sbin/python (system Python 3.13.7)
2. The virtualenv’s Python binary (./venv/bin/python) was correctly linked to Python 3.12.12
3. System Python’s sys.path didn’t include the virtualenv’s site-packages directory
4. The source .venv/bin/activate command wasn’t properly updating the shell’s PATH to prioritize the virtualenv’s Python

Solution

Correct Virtualenv Activation

Explicitly used the virtualenv’s Python binary (./venv/bin/python) instead of relying on system Python, ensuring proper path resolution:

./venv/bin/python main.py

Path Verification

Confirmed the virtualenv’s Python had correct paths:

./venv/bin/python -c "import sys; print('\n'.join(sys.path))"

This showed the virtualenv’s site-packages (/home/user/.local/lib/python3.12/site-packages) was properly included.

Import Correction

Fixed MoviePy import syntax from:

from moviepy.editor import VideoClip, AudioFileClip

to:

from moviepy import VideoClip, AudioFileClip

TTS Pipeline Implementation

Implemented proper Kokoro pipeline initialization:

from kokoro.pipeline import KPipeline
pipeline = KPipeline(lang_code='a')  # American English
generator = pipeline(text, voice='am_echo')

The solution required explicit virtualenv Python usage and careful import path management to resolve the version mismatch issues while maintaining all package dependencies.