Free · Private · No upload Updated June 2026

Find out if your interview English will cost you the job — in 60 seconds

InterviewVoice IRI Score is the only interview prep tool that gives non-native English speakers a single, formula-transparent composite score — computed live in your browser against named peer-reviewed speech-science benchmarks. No AI black box. No upload. No account.

Start Free IRI Test
Benchmarks: Griffiths (1990) 127–188 WPM comprehension range · Tauroza & Allison (1990) 145 WPM native lecturer average · Gósy et al. (2020) 2.1 fillers/min non-native baseline
Example output — live tool produces your real score
78 IRI
Q2 · Behavioral
"Describe a challenge you faced at work and how you overcame it."
Pace Legibility
85
148 WPM
Filler Control
72
2.8 fillers/min
STAR Structure
67
Opening ✓ Evidence ✓ Closing ✗
I have been working in data engineering for three years. Um, the biggest challenge was when our pipeline failed. Specifically, we had to rebuild the ETL process — like, completely from scratch. Basically I led a team of four engineers…

Choose your interview question

Pick one to practice now, or cycle through all five for a full session. Your answer is processed entirely in your browser — nothing is sent to any server.

Tell me about yourself and your background.
Ready
0:00
Elapsed
— WPM
Live pace
0
Fillers
Click the mic to start your 60-second answer
Your spoken words will appear here in real time as you speak…
—
IRI SCORE

Computing…

Question: —

Sub-scores

Pace Legibility
—
—
Filler Control
—
—
STAR Structure
—
—

IRI Formula

IRI = 0.35 × Pace + 0.35 × FillerControl + 0.30 × STAR
—
Weights reflect relative impact on interview outcome per speech-science literature. Pace and filler control each carry 35%; structure carries 30%.

Annotated transcript

—
Filler word
STAR Opening signal
STAR Evidence phrase
STAR Closing signal
How it works

Three research-backed sub-scores. One defensible number.

The IRI is not a black-box AI rating. It's a transparent weighted formula computed from your own transcript against named, peer-reviewed benchmarks — so you can explain every point to a mentor.

Pace Legibility Score (35% weight)

Optimal band: 130–160 WPM. Each 10 WPM outside the band deducts 15 points (floor 0). Derived from Griffiths (1990) 127–188 WPM comprehension range and Tauroza & Allison (1990) 145 WPM native-lecturer centroid.

Score = 100 − 15 × max(0, ceil((|WPM−145|−15)/10))
Griffiths 1990 (RELC Journal) ↗ · Tauroza & Allison 1990 ↗

Filler Control Score (35% weight)

Counts 'um', 'uh', 'like', 'you know', 'basically', 'literally', 'right' per minute. Maps to a percentile using the Gósy et al. (2020) non-native speaker baseline: mean 2.1/min, SD 1.4/min (normal distribution). Score = 100 − percentile.

p = Φ((user_fpm − 2.1) / 1.4) × 100
FillerControl = 100 − p
Gósy et al. 2020 (J. Fluency Disorders) ↗

STAR Structure Score (30% weight)

Detects three structural signals in your transcript: Opening ('I am', 'My background', 'I have', 'I work'), Evidence ('for example', 'specifically', 'as a result', 'which meant'), Closing ('learned', 'ultimately', 'outcome', 'key takeaway'). Each detected = +33 pts.

STAR = (opening ? 33 : 0) + (evidence ? 33 : 0) + (closing ? 34 : 0)
Based on STAR interview framework (Situation, Task, Action, Result) — standard in behavioral interviewing literature.
IRI Composite Formula
IRI = 0.35 × PaceScore + 0.35 × FillerControl + 0.30 × STARScore
Score bands: 0–49 Needs work · 50–74 Developing · 75–100 Interview-ready
Research benchmarks

The data behind your score

Every IRI number traces to a specific published figure. No invented statistics.

Metric Benchmark value Source IRI application
Optimal speech rate 127–188 WPM comprehension range Griffiths (1990), RELC Journal Lower-risk zone 130–160 WPM used as IRI sweet spot
Native lecturer pace 145 WPM average Tauroza & Allison (1990) Centroid of optimal band; 0-penalty midpoint
Non-native filler rate Mean 2.1/min, SD 1.4/min Gósy et al. (2020), J. Fluency Disorders Percentile baseline for Filler Control sub-score
Comprehension drop Significant above 200 WPM Griffiths (1990) Upper hard limit; scores 0 above 200 WPM
Filler threshold (bottom quartile) ~4–5 fillers/min (≈75th percentile of Gósy distribution) Derived from Gósy et al. (2020) normal distribution Filler Control score <35 triggers "high filler" warning

Disclosure: The Gósy et al. (2020) figure of 2.1 disfluencies/min refers to non-native speakers in formal speaking contexts. The Griffiths (1990) and Tauroza & Allison (1990) rates are from academic lecture settings. Interview speech may differ slightly. The IRI is an evidence-informed estimate, not a clinical assessment.

Recommended tools

Go deeper on interview prep

Tools our users find genuinely useful alongside the IRI Score.

Grammarly Premium
After you get your IRI transcript, paste it into Grammarly to catch grammar patterns that may undermine your written follow-up emails and LinkedIn messages. The tone detector flags overly formal or informal register — useful for non-native speakers calibrating professional English.
Try Grammarly →
Affiliate link — we may earn a commission if you subscribe. We recommend Grammarly because it has a free tier and a genuine non-native English use case.
Speechify
Listen to your interview answers read back at different speeds (0.8× to 1.5×) to train your ear for the 130–160 WPM target range. Speechify's text-to-speech engine is accurate enough to model natural interview pacing — useful for calibrating what 145 WPM actually sounds like.
Try Speechify →
Affiliate link — we may earn a commission if you subscribe. Recommended for pace calibration training.
LinkedIn Premium Career
Once your IRI score is above 75, LinkedIn Premium's Interview Prep feature adds human-reviewed mock interviews and role-specific question banks. Pair it with your IRI baseline to track improvement over a structured 2-week sprint before your interview date.
Try LinkedIn Premium →
Affiliate link — we may earn a commission if you subscribe. LinkedIn Premium is the most direct path to recruiter visibility after improving your spoken English.
FAQ

Common questions

InterviewVoice IRI Score is free, runs entirely in your browser, and gives you a live WPM counter plus a composite Interview Readiness Index benchmarked against peer-reviewed speech-science research (Griffiths 1990; Tauroza & Allison 1990). No account, no upload, no server. Works on Chrome, Edge, and Safari.
Gósy et al. (2020, Journal of Fluency Disorders) found fluent non-native speakers average 2.1 disfluencies per minute in formal speaking contexts (SD 1.4). Above ~4–5 fillers/min places you in the bottom quartile of that distribution. The IRI counts your fillers live from your browser transcript and maps them to this published percentile — so you see exactly where you stand relative to the research baseline, not a vague "too many."
Yes — InterviewVoice uses the browser's built-in Web Speech API. Your audio is transcribed locally by your browser and never leaves your device. The IRI composite score (pace + filler density + STAR structure) is computed entirely in JavaScript on your machine. The formula is published openly on this page so you can verify every calculation.
Griffiths (1990, RELC Journal) found non-native listeners comprehend speech best at 127–188 WPM, with comprehension dropping significantly above 200 WPM. Tauroza & Allison (1990) measured native English lecturers at 145 WPM on average. The IRI optimal band is 130–160 WPM — the lower-risk zone of the Griffiths range centered on the Tauroza & Allison native-lecturer average. Below 100 WPM can signal hesitation; above 180 WPM risks comprehension loss.
InterviewVoice IRI Score was built specifically for non-native English speakers preparing for high-stakes interviews (H1-B, OPT, international roles). It scores filler density against a non-native speaker baseline (not a native-speaker standard), checks STAR structure (Opening, Evidence, Closing), and generates a shareable result card you can screenshot and review with a mentor or immigration advisor. It's free, private, and requires no account.
The IRI measures three specific, measurable signals — not overall quality, confidence, or content accuracy. A 60-second answer that is substantively excellent but spoken at 200 WPM with 6 fillers/min and no STAR structure keywords will score in the 40s. That's the point: the IRI isolates the delivery mechanics that research links to comprehension and perceived fluency, so you can fix them separately from content. Use it as a diagnostic, not a verdict.
The tool works on iOS Safari and Chrome for Android. The Web Speech API is supported on both. Firefox on any platform does not support Web Speech API — use Chrome or Safari instead. For best transcript accuracy, use a quiet environment and speak clearly into your device's built-in microphone.