👁️ Howdy — IR Facial Recognition on Ubuntu 24.04

Windows Hello for Linux — Complete Setup Guide

Howdy brings Windows Hello-style facial recognition to Linux. It uses your laptop's built-in IR (Infrared) camera — the small dark oval on the webcam strip — to authenticate you at login, sudo prompts, and more. No password needed, just look at the camera.

⚠️ Disclaimer: This is not created by me — just a setup guide by a clueless kid learning how to set up Howdy on Ubuntu. Use at your own risk.

• ✔️ Works with Howdy
• ✔️ Easier to set up
• ❌ Can be fooled by a photo of your face
• ❌ Affected by lighting changes (too dark = fails)
• ❌ Less accurate overall

• ✔️ Can't be fooled by photos
• ✔️ Works in complete darkness (IR emitter lights your face invisibly)
• ✔️ More accurate and faster recognition
• ✔️ What Windows Hello uses

💡 Recommendation: Use the IR camera for best results. It's more secure and reliable.

Ubuntu 24.04 uses PEP 668 which blocks system-wide pip installs by default.
Always use --break-system-packages flag:

sudo pip install --break-system-packages \
  dlib \
  face_recognition \
  opencv-python \
  ffmpeg-python \
  numpy --ignore-installed

⚠️ If numpy throws a conflict error, add --ignore-installed — it's already installed via apt and works fine.

sudo add-apt-repository ppa:boltgolt/howdy
sudo apt update
sudo apt install howdy

💡 If install gets interrupted, fix with:

sudo dpkg -i --force-all /var/cache/apt/archives/howdy_2.6.1_all.deb
sudo dpkg --configure -a

Howdy 2.6.1 on Ubuntu installs as a PAM module only — no binary in PATH.
Create an alias to use it like a normal command:

echo 'alias howdy="sudo python3 /usr/lib/security/howdy/cli.py"' >> ~/.bashrc
source ~/.bashrc

Now howdy add, howdy config, howdy list all work.

Howdy's pam.py uses the Python 2 module name ConfigParser which doesn't exist in Python 3:

sudo sed -i 's/import ConfigParser/import configparser as ConfigParser/g' \
  /usr/lib/security/howdy/pam.py

v4l2-ctl --list-devices

Test each device visually — IR camera will look grey/monochrome:

ffplay /dev/video0    # regular webcam — shows color
ffplay /dev/video2    # IR camera — shows grey/infrared

Your IR camera is usually /dev/video2.

howdy config

Change these settings:

[video]
device_path = /dev/video2      # your IR camera device
certainty = 4.5                # lower = stricter (1-10)
timeout = 10                   # seconds before giving up
recording_plugin = opencv      # more stable than ffmpeg for IR
force_mjpeg = true             # fixes IR frame decoding issues
dark_threshold = 50            # handles IR lighting

[snapshots]
capture_failed = false         # disable to avoid permission errors
capture_successful = false

howdy add

• Look directly at the IR camera (dark oval — not the regular webcam)
• Enter a label like zang when prompted
• Stay still and well-lit while frames are captured

Check if Howdy is already in PAM:

cat /etc/pam.d/common-auth | grep howdy

If empty (nothing returned), add it manually:

sudo nano /etc/pam.d/common-auth

Add this as the very first line:

auth sufficient /usr/lib/security/pam_python.so /usr/lib/security/howdy/pam.py

💡 Find the correct pam_python.so path with:

find / -name "pam_python.so" 2>/dev/null

Suppress OpenCV GStreamer warnings:

echo 'export OPENCV_LOG_LEVEL=ERROR' >> ~/.bashrc
source ~/.bashrc

Fix Python deprecation warning in compare.py:

sudo sed -i 's/datetime.datetime.utcnow()/datetime.datetime.now(datetime.timezone.utc)/g' \
  /usr/lib/security/howdy/compare.py

Fix snapshot permission error:

sudo chmod 777 /usr/lib/security/howdy/snapshots

sudo -k
sudo ls

Look at the IR camera. You should see:

Identified face as zang

howdy add        # add a new face model
howdy list       # list saved models
howdy remove     # remove a model
howdy enable     # enable Howdy
howdy disable    # disable Howdy temporarily
howdy config     # edit configuration

This entire setup — every error, every fix, every config tweak — was debugged live with Claude by Anthropic.

  ╔══════════════════════════════════════════╗
  ║                                          ║
  ║        ✦  Claude by Anthropic  ✦        ║
  ║                                          ║
  ║     AI that guided this setup from       ║
  ║     first error to final working         ║
  ║     facial authentication — live.        ║
  ║                                          ║
  ║          https://claude.ai               ║
  ║                                          ║
  ╚══════════════════════════════════════════╝

Made with ❤️ on Ubuntu 24.04  •  Debugged with Claude (Anthropic)