Questions, answered.

Lyra is a practice room for job interviews. You speak with an AI persona who plays a hiring manager, recruiter, or panel of interviewers — they ask questions, follow up, and react in real time. Afterwards, you get a calibrated report on what worked and what didn't. It's not a replacement for the real thing; it's a low-stakes way to get the reps in.

No. Lyra runs in the browser — Chrome, Safari, Firefox, and Arc are all supported on desktop and mobile. We use WebRTC for the voice loop, so the only real requirement is a working microphone and a stable connection.

If you're on a corporate device that blocks WebRTC, sessions won't start. Switch to a personal browser profile or check with your IT team.

Most people run 20–35 minute sessions. You can stop at any time and still get a partial report. Panel interviews tend to run longer (45–60 min) because each persona takes a turn.

Yes. The free tier includes three full sessions with feedback reports, no credit card required. After that you'll need a plan to keep going — see pricing for details.

Behavioural, system design, product sense, case studies, and recruiter screens. We don't yet cover language-specific coding interviews — for now we recommend pairing Lyra with a tool like LeetCode for the algorithmic work, and using us for everything around it.

Nine times out of ten it's one of these:

• Browser permission denied — click the lock icon next to the URL and toggle Microphone to Allow.
• The wrong input is selected — use the device picker in the pre-call screen to switch to your headset.
• Another tab is holding the mic (Zoom, Meet, Discord). Close it and refresh.

Just talk over them. Since v2 we use real-time audio with barge-in detection, so the persona will stop speaking when you start — same as a real interview. This is a learned skill; if you're hesitant about interrupting, it's worth practicing.

Yes — in practice mode you can redo any question as many times as you like. In a full session, say “let me start over”, the persona will rephrase from your report.

Each persona has a role, an interviewing style, and a set of question patterns they tend to favour. Maya is a design hiring manager who pushes on craft and trade-offs. Ben is a recruiter who runs tight, warm screening calls. Aki is a director-level engineer who lingers on ambiguity. Their voices are different. Their pacing is different. Choose based on the loop you're prepping for.

Each answer is scored across seven dimensions — relevance to the question, relevance to the job description, seniority signal, clarity, confidence, impact, and ownership. Each dimension is rated 1–5, where 3 means “solid and competent for this role”. A weighted score is computed from those seven dimensions and shown alongside a one-sentence note pointing at a specific moment in your transcript. The session-level score is the weighted average across all questions, but the per-dimension breakdown is usually more useful than the headline number.

Your report covers every question you answered. For each one you get: a weighted score across seven dimensions, a STAR-method breakdown for behavioural questions, a list of strengths and areas to improve, and — for lower-scoring answers — a rewritten example showing what a stronger answer could look like. There's also an overall summary across the full session.

Usually a few seconds after your session ends. You can wait on the page — the report appears automatically when it's ready.

Yes — your session history lives in the dashboard. Each entry opens the full transcript and feedback report for that session.

Four options: Free (one full session, no credit card needed), Extra Token ($4.99 one-time, one more session, never expires), Pro ($19/month — 24,000 credits, roughly 4–5 full sessions or 60 minutes of voice practice), and Premium ($49/month — 64,000 credits, roughly 10–12 full sessions or 160 minutes). Pro and Premium are monthly subscriptions; Extra Token is a one-time top-up you can use any time. All options include feedback reports.

Settings → Billing → Cancel subscription. One click, no friction, no exit interview. Your access continues to the end of the current billing period.

A one-time $4.99 purchase that adds one full session (6,000 credits, up to 15 minutes of voice practice) to your account. It doesn't expire and doesn't require a subscription — useful if you want one more session without committing to a monthly plan.

No. Your transcripts and audio are never used to train language models, ours or anyone else's. The only data we look at in aggregate is anonymised quality signals — things like “how often does the report citation match the right moment in the transcript”. You can opt out of even that in Settings.

Audio is deleted within 30 days of your session. Transcripts and reports stay in your account until you delete them — they're yours. You can wipe everything from Settings → Privacy with a single confirmation.

Stored encrypted at rest, used only to personalise your sessions, never shared with anyone (including recruiters — we don't have a recruiter side). Delete it any time from Settings.