A 1000-point score for speaking practice
The Wellspoken Index is a 1000-point score that summarizes a recording across six dimensions: Structure, Conciseness, Confidence, Pronunciation, Filler Rate, and Pace.
Structure
How clearly your ideas are organized, sequenced, and supported.
LLM evaluation
Conciseness
How efficiently you get to the point and avoid avoidable repetition.
LLM evaluation
Confidence
How steady, assured, and direct your delivery sounds in the recording.
Transcript and delivery-pattern evaluation
Pronunciation
How clearly words and sounds are articulated so listeners can follow.
Deterministic speech scoring with AI feedback
Filler Rate
How often fillers such as um, uh, like, you know, right, and so interrupt your response.
Deterministic formula with AI feedback
Pace
How well speaking speed and variation support clarity throughout the recording.
Deterministic formula with AI feedback
| Dimension | Weight | Method | Sub-metrics |
|---|---|---|---|
| Structure | 250 pts | LLM evaluation | Logical Sequence, Transitions, Signposting, Opening Quality, Closing Quality |
| Conciseness | 200 pts | LLM evaluation | Word Choice, Word Economy, Sentence Length |
| Confidence | 150 pts | Transcript and delivery-pattern evaluation | Hedging Frequency, Uptalk, Assertiveness |
| Pronunciation | 150 pts | Deterministic speech scoring with AI feedback | Pronunciation Clarity |
| Filler Rate | 150 pts | Deterministic formula with AI feedback | Filler Frequency |
| Pace | 100 pts | Deterministic formula with AI feedback | Pause Timing, Words per Minute |
What the score is for
The Index is a Wellspoken coaching score for practice sessions. It is designed to make feedback easier to scan, show which dimension deserves attention next, and help you review patterns across repeated recordings.
How Wellspoken measures it
Wellspoken combines transcript, timing, and audio-informed analysis where available to generate the six dimension scores. Read the score with the transcript, audio playback, and written feedback for the full context of a session.
Transcript reasoning
Structure and Conciseness use the transcript to evaluate how ideas are ordered, connected, and expressed. These dimensions need judgment about meaning, so they use LLM evaluation.
Delivery-pattern analysis
Confidence uses observable speech patterns such as hedging, uptalk, sentence endings, pacing context, and assertive phrasing.
Deterministic measurements
Pronunciation, Filler Rate, and Pace rely on measurable speech data. Pronunciation uses speech scoring, Filler Rate uses filler frequency, and Pace combines pause timing with active speaking speed.
How pace is interpreted
Pace uses active speaking speed, which excludes pauses, alongside pause timing. That means the number can be higher than a simple overall words-per-minute count. The goal is controlled rhythm: enough pace to sound energetic, enough pausing to make ideas easy to follow.
How to improve it
Start with your lowest dimension, practice that skill in a short drill, and compare later sessions against earlier ones. The dimension breakdown usually matters more than the total number because it tells you what to work on next.
Learn more
For a deeper walkthrough of how the score works, read the full guide to the Wellspoken Index.
Download Wellspoken