I want to add my answer, even though many of the key points have already been covered by @sahasrara62, @aSteve, and, of course, @Daweo. I’ll compile all the information into a single, structured explanation, along with some of my own thoughts.

As @aSteve mentioned, if you install numpy under Python 3.12, Python 3.13 won’t see it, and the same goes for pyaudio in reverse. This is why VS Code’s Pylance reports “could not be resolved” — it’s simply looking at the wrong Python environment.

Following @Daweo’s sugggestion, you can explicitly install packages for a particular interpreter by running:

python3.12 -m pip install numpy 
python3.13 -m pip install pyaudio 

This ensures that the correct pip tied to each Python version is used, eliminating any ambiguity.

However, as @sahasrara62 pointed out, a much better long-term approach is to create a virtual environment tied to a single Python version. This avoids the “wonky” setups you described and is the industry standard for managing Python dependencies. It keeps everything isolated and predictable.

For example, to create and activate a virtual environment with Python 3.12, you can run:

python3.12 -m venv myenv source myenv/bin/activate
# On Windows it would look like thuis: myenv\Scripts\activate 

All your packages will live inside myenv, completely separate from other Python installations and projects. I highly recommend using this approach.

Not to forget. If you’re working in VS Code, make sure it’s set to use this virtual environment. Open the command palette (Ctrl+Shift+P), type Python: Select Interpreter, and choose the one that points to myenv. This will also resolve the missing import errors reported by Pylance.