Painting portraits from live reference versus painting them from photo reference produces different results and requires different skills. I've done both extensively and want to compare the experiences.

Live reference: you observe the subject directly, which trains spatial understanding and forces fast decisions. The model moves slightly between glances, so you develop the ability to store and reconstruct form. The quality of your observation is the ceiling.

Photo reference: the subject is frozen, which allows slower, more careful analysis. But photos lie in specific ways — lens distortion, flattened depth, uneven color temperature from the light source. Painting from a photo without knowing these limitations produces work that looks like a photo filter, not a painting.

The hybrid approach I recommend: use live reference for gesture and quick studies to build spatial intelligence. Use photo reference for finished work with awareness of the photo's limitations. The live practice calibrates your eye so you can correct for the photo's distortions.