Understanding the Results

The structure of a result, the vital keys, when each value is available and how to use the confidence scores are the same for all VitalLens clients and described in Understanding Results. This page covers what is specific to the Python client.

One Result per Face

Processing a video returns a Python list with one result per detected face. If two people are visible, the list has two entries.

results = vl(video)
# Type: List[dict]
hr = results[0]['vitals']['heart_rate']['value']

Waveform data, confidences and timestamps are returned as lists; scalar values are Python floats.

Besides face, vitals, waveforms, message, model_used (VitalLens API only) and core_version, each result contains:

Key	Contents
`rolling_vitals`	Vitals over time, see below.
`fps`	The frame rate the video was processed at.
`n`	The number of frames.
`time`	The timestamp of each frame in seconds.

Rolling Vitals

When processing a video file, the client also estimates each vital over time, as one value per frame from a trailing window (the window is listed under Availability). They are under rolling_vitals, with the same keys as vitals:

rolling_hr = results[0]['rolling_vitals']['heart_rate']['data']  # one value per frame

Rolling vitals are on by default. To skip them:

vl = VitalLens(estimate_rolling_vitals=False)

File Export

The client also saves the results to a timestamped JSON file (e.g., vitallens_20260203_120642.json) in your current directory. To disable this:

vl = VitalLens(export_to_json=False)