VoxElevation

Understanding Your Scores

What the scores mean, how confidence levels work, and how AI generates your results.

The scoring scale

Scores run from 1.0 to 5.0. A 5.0 means excellent performance in that dimension; a 1.0 indicates significant room for improvement. Scores around 3.0–3.5 are typical for solid delivery with specific areas to develop. Your overall score is a weighted average across all measured dimensions.

Score categories

VoxElevation scores your delivery across several dimensions including energy and vocal dynamics, pacing and flow, physical presence and movement, structural clarity, and inferred audience engagement. Not all categories can always be measured — see "Unmeasured categories" below.

Are scores generated by AI?

Yes. Scores are generated by analyzing your video using transcription, motion tracking, and audio processing. The system measures signals like speaking rate variability, volume dynamics, physical movement patterns, and verbal structure. Some dimensions are measured directly; others are inferred from related signals.

Confidence levels

Each score has a confidence level indicating how reliably it was derived: • High — derived directly from clear, measured data. Full weight in your overall score. • Medium — derived from proxy signals with reasonable accuracy. 0.7× weight. • Low — weak proxy estimate; treat as directional only. 0.4× weight. • Human — supplied via a self-assessment you filled in. Full weight. • Unmeasured — data wasn't available to score this dimension. Excluded from overall score.

What does "Inferred" mean?

An inferred score is estimated from indirect signals rather than measured directly. For example, "audience response" cannot be measured from your video alone — there's no room microphone. Instead, it's estimated from delivery patterns that typically correlate with engagement (energy variation, pacing, eye contact proxies). Inferred scores are clearly labeled so you know they're estimates.

Why are some categories unmeasured?

Some scoring dimensions require data that isn't always available from the uploaded video. For instance, if audio quality is very poor, certain speech metrics can't be reliably extracted. Unmeasured categories are shown explicitly so you know they weren't forgotten — just not scorable from the available data.

Accessibility signals

Accessibility signals show how your delivery patterns may affect diverse listeners — people who process, hear, or engage differently. These are informational observations derived from your delivery data, not scores. They highlight areas like pace, vocabulary complexity, and repetition that can improve reach across your audience.

Self-assessment scores

After reviewing your report, you can fill in a reflection form with your own self-assessment across several dimensions. These become "Human" confidence scores that are included in your overall weighted score and tracked over time alongside your AI-generated results.