Everything you wanted to know.

Expression and tone reads are impressions, not measurements of what you felt. The model estimates how your face and voice are likely to read to a viewer, the way an interviewer forms a gut sense of you. When you calibrate, close calls between similar expressions are settled the way your own face shows them — a clearly serious expression still reads as serious. Calibration is available on paid plans.

What this means in practice: if Veracue tells you your face read "concerned" while your voice read "neutral", treat that as how the moment likely landed on the other side of the table — worth reviewing at the timestamp, not a verdict on what you felt. The pace, filler-word and transcript figures are measured directly from the recording, and you can verify each of them yourself. Eye contact is scored from where your gaze sat, frame by frame.

No — and the score reflects it. The substance of your answer — whether you actually addressed the question, with specifics, in a structure an interviewer can follow (including STAR on behavioural questions) — carries the largest share of your overall score. Delivery signals like pace, fillers, eye contact and expression are analysed alongside it, in the same answer. Polished delivery can't carry an empty answer here.

Content and delivery in one honest analysis. Most tools do one or the other: transcript graders mark your answer like an essay, and body-language tools ignore what you said. Veracue scores substance first and delivery alongside it, names the patterns it sees (for example, gaze that reads as reading from notes), and tells you the one thing to fix next. And it's a practice tool, full stop — never a live-interview copilot. We don't do stealth assistance in real interviews. For a feature-by-feature table against Yoodli, Big Interview and Huru, see how we compare.

It's a weighted blend of three scores: “What you said” 50%, “How you said it” 29% and “First impression” 21%. What you said — did you actually address the question, with specifics, in a structure an interviewer can follow, including STAR on behavioural questions — carries the largest share. There's also a cap: your overall score can sit at most 25 points above your score for what you said, so polished delivery can never carry an empty answer.

Read 80+ as an answer worth locking in, 60–79 as solid with clear lifts available, and below 60 as a signal to rebuild the answer around one real example — your results name the single thing to fix first.

Real research, with every source named on our research page and the contested parts said so. The methodology draws on five decades of communication and social-psychology research — Ekman's universal facial expressions, Mehrabian's early work on non-verbal communication (whose famous percentages we deliberately don't quote — they're far narrower than pop culture suggests), and more recent work on vocal arousal and emotional congruence (Newcombe & Ashkanasy, 2002; Pentland, 2008).

You can read the full bibliography on our Latest Research page. Where research disagrees, we've noted it transparently.

They aren't stored, unless you choose to keep them. Your answers are analysed on our servers and the recording is deleted within seconds of the analysis completing; Keep my recordings holds a watch-back copy for 90 days (see the Privacy Policy, section 2.8). A failed analysis is held for up to 24 hours for debugging, then automatically deleted. Otherwise we keep only your transcripts, scores and feedback so you can track progress.

You can delete everything from Settings → Data & Privacy at any time. We do not train shared models on your answers without explicit opt-in.

Partly — and for interview preparation, that's fine. The reads are impressions of how you come across, not a lie detector. If you can keep a steady pace and eye contact for 90 seconds while saying something substantive, that isn't fooling the model — that's the skill working.

No. Any laptop webcam from the last 5 years is fine, and so is a phone’s front camera — stand the phone up at eye level rather than holding it. What matters far more is lighting (face a window, not your back to one) and camera height (eye level, not staring up your nose). See Tips for a five-minute setup that removes the most common causes of a poor read.

60 to 120 seconds is the sweet spot. Shorter than 60 and there isn't enough signal to grade reliably — the confidence intervals widen significantly. Longer than 2 minutes and most interviewers have already decided.

If a question genuinely needs more, 90 seconds is usually the right move.

No. Veracue is a personal practice tool. We have no enterprise admin features, no manager dashboards, and no integrations with HR or recruiting platforms. Your answers are yours.

English is fully supported, including Australian, British, American, Indian and Irish accents. The expression and tone reads are language-agnostic by construction — they read face and voice, not words — but transcription, pace and substance scoring currently run in English.

If your accent is being mistranscribed, please report it from the answer's "Report a problem" link — those reports go straight to the fix list.

No, and we'd be uncomfortable with anyone who claims otherwise. Hiring decisions involve role fit, technical depth, references, salary, timing, and luck. Veracue covers one part of that:the answer itself, and how it landed — whether you addressed the question with specifics and structure, and how your delivery read while you did. Both are consistently underweighted by candidates and consistently decisive for interviewers — but they aren't everything.

No — no practice tool can. Veracue gives you honest feedback on how your answers land, so you walk in prepared. What you do with that is what gets you hired.

Yes. You get 2 free practice questions, each with the full feedback on what you said, how you said it and how you came across. Pro ($29 USD a month, or $42 AUD a month in Australia) gives you 25 practice questions each billing period, Premium ($49 USD a month, or $70 AUD a month in Australia) 80, and Interview Sprint ($39 USD, or $56 AUD in Australia, one payment) 50, valid 60 days from purchase. Prices are shown and charged in your local currency where we offer one (currently AUD and USD); otherwise in USD.

A practice question is one question you answer and get feedback on — it’s what your plan counts. A practice session is one sitting: usually three questions, and you can choose up to five.

Solo practice is what Veracue is built for — we want you alone with the camera, because that mirrors a real video interview. If you want a coach's eyes on your results, walk them through the results page together — there's no share link yet.

Only if you choose to. The default is no. Two choices lead to it. If you tick the model-improvement box when you calibrate, your calibration photos and readings are kept against your account — not anonymised — and used to train and improve our models. If you turn on Keep my recordings, the face images, movement measurements and clips kept from your recordings are used to train and test our face-reading models (Privacy Policy, section 3). You can withdraw from either any time in Settings, which removes what you gave from all future training; models already trained before you withdraw cannot be untrained. Never sold, never shared with third parties.

Because you swing between answers. Sleep, caffeine, the question difficulty, and how recently you ate all measurably move your delivery. The longitudinal trend across many answers is the signal — any single answer is noise.

Yes, at specific moments we measured, each with the words you were saying at the time. A smile that stayed up through a difficult part of your answer, a brow that lifted, a look away from the camera: you can watch each moment back. We never turn it into an overall score, and we never say what a look meant about you: the read is good enough to point you at a moment worth a look, not to grade you.

Imperfectly, and we'd rather say so than pretend otherwise. Expression models carry known biases from their training data, which is why Veracue treats their output as impressions rather than facts, weights your words above your face, and offers calibration, which settles close calls between similar expressions the way your own face shows them. It adjusts reads; it doesn't replace the population model or re-read a clear expression.

This is the hardest open question in the space and we won't pretend we've solved it. The current models are trained on neurotypical baselines; if your natural eye contact pattern or facial expressiveness differs from that baseline, the absolute scores will systematically underestimate you. The longitudinal trend (you vs you) remains valid; the absolute number does not.

We want to be precise about what calibration does. The underlying model stays population-trained. Calibration only acts on close calls — moments where the model's two leading reads are nearly level — and settles them the way your reference photos show your face. It never overrides a clear read, so a serious or flat resting expression still registers in your delivery reads. Calibration is available on paid plans, and it's optional, consent-gated, and deletable any time from Settings.

Practise the real skill: answering questions out loud, under time pressure, and reviewing what actually came out. Build a bank of six to eight real examples, answer questions on camera, read the transcript back, and fix one thing per session — substance first (a specific example, a clear structure, a real result), then delivery signals like pace, fillers and eye contact.

Focused reps with honest review beat hours of re-reading notes. Our behavioural questions guide covers the structure side in depth.

Rambling is a structure problem, not a talking problem. Decide your ending before you start: answer behavioural questions with one example in STAR shape, and stop when you reach the result. Aim for 60 to 120 seconds. Recording yourself and reading the transcript shows exactly where you loop or restart — usually at the point where the answer had no planned ending. See how to structure answers that land.

Recreate the real conditions: camera at eye level, face a window for light, look at the lens, and answer within a time limit on the first take. One-way interviews give you no interviewer reactions to lean on, so structure and a steady delivery carry more of the impression. Record practice answers, review the transcript and delivery, and rehearse until a 60-to-120-second structured answer feels normal — that's exactly the loop a Veracue session runs.

Yes, on every plan. Every question has a Listen button — on the review screen and again when you're in front of the camera — so you can hear it read in an interviewer's voice instead of reading it yourself. Nothing plays unless you press it. In simulation mode the question is asked aloud automatically, and recording starts a moment after the voice finishes, so the voice is never in your recording. The text stays on screen either way.

A practice session that feels like the real thing. You pick your questions as usual, then choose simulation mode instead of standard practice. Each question is asked aloud (it stays on screen too), recording starts on its own a moment after the voice finishes, and the moment you stop, the next question is there — no pause, nothing to press between answers. If a question's audio can't load, it shows with a short read count instead. Standard practice records the same way but you start each answer yourself, with a moment to think in between. In both, your answers analyse in the background while you keep going, and when the last one is in you get a summary of the whole session with a link to the full feedback for each answer.

Two things to know before you choose it. Each answer uses one practice question, exactly as standard practice does, and the picker tells you how many before you start. And there is no feedback until the end in either mode — simulation just takes away the pause. Simulation mode is included with Pro, Interview Sprint and Premium, and needs a desktop browser.

The closest thing to the real thing. Tick “Do this as a formal interview” under the job description, choose three or five questions, and you never see a question before it is asked. We choose them for the role, leaving out anything you have answered before. The camera is already recording when each question is read aloud — the way you listen is part of your first impression, as it is with a real interviewer — and the text appears only once the question has been asked. You answer, and the next question follows the moment you stop. Nothing you hear from the interviewer is transcribed or counted against you: the analysis starts where your answer does.

Before it starts you see the industry we matched, the count, and the cost — each answer uses one practice question, as in any session — and nothing else. There is no feedback until the end, and no way to re-roll a question you haven’t heard yet. If you leave part-way, the answers you have already stopped still count; the one you are on does not. Included with Pro, Interview Sprint and Premium, on a desktop browser.

Practice for something that isn’t an interview: a pitch, a speech or a presentation. You say what you’re practising and what would make it go well, we write it up as a goal and three to five criteria, you change any of them and confirm, then you record. You get a score against your own criteria and the usual feedback on how you said it. A session uses one practice question, can run for up to three minutes, and stays out of your interview scores and trends. Included with Premium only.

Yes. The question bank covers graduate and entry-level roles across 43 industries, from clerkships and graduate programs to first nursing, teaching and engineering jobs, and a resume on file tells the selector your level. Behavioural questions, which graduate interviews lean on, are scored on whether you gave a real example with your own part in it.

Yes. A behavioural question is scored on what happened, what you did and how it ended, as well as on whether you answered the question with real detail. Talking points for a behavioural question walk you through exactly that, and the feedback tells you which part never arrived.

Yes. Pick intermediate for talking points or expert for the question alone, on any plan. Answers given in your own words are the ones your progress is tracked on; a script is there for when you have no idea where to start.

Yes, in the phone’s browser: Safari on iPhone, Chrome on Android. Stand the phone up at eye level rather than holding it. Simulation mode, formal interviews and watching your recording back need a computer, and in-app browsers (Instagram, TikTok) cannot record.

No. Veracue is made in Australia and works anywhere in English; the question bank and the scoring are not tied to one country. Prices are in US dollars, or Australian dollars for a visitor in Australia.

Yoodli runs a live roleplay with follow-up questions based on your answers; Veracue does not. Veracue writes you a script in your own words, scores what you said against a rubric, and tells you what your face and voice did at which words, as moments you can watch back. The full table is at how we compare.

Two answers with full Pro feedback: the three scores, the coaching, the face, eye contact and voice cards, and the improved-script button. No card is needed, and the second answer is the same question again so you can see your score move.

No, unless you turn on Keep my recordings. Your recording is analysed on our servers and deleted within seconds of the analysis completing; a failed analysis is held for up to 24 hours for debugging, then deleted. We keep your transcript, scores, feedback and a written list of moments (never an image of your face), as the Privacy Policy sets out.

No. It measures eye contact, pace, filler words, your expression at the words you were saying, and what you said; nothing it shows is a verdict on what you felt or who you are.

Settings → Subscription → Cancel. One click. No call. No retention flow. Your access runs to the end of the billing period you have paid for. There are no change-of-mind refunds — your Australian Consumer Law rights are unaffected.