Knowing the answer is not the same as saying it out loud in nine minutes.
VoiceOSCE is spoken rehearsal for the Part B circuit. You talk through a station with an AI examiner, you get your own words back beside the model answer, and you see exactly which marking points you missed.
No card required for the trial. Microphone permission is requested before anything listens.
You apologised within your first reply. That is the step that de-escalates this station.
The loop, four moves
Choose a station, rehearse it out loud, read the debrief against the mark scheme, then repeat the one thing you got wrong.
- 01
Choose a station
Filter by category, by a weak area from your last debrief, or pick up the unfinished attempt waiting on your dashboard.
- 02
Read the scenario, then start the clock
The examiner reads the briefing aloud and it stays on screen for the whole station, as it does on the day.
- 03
Say your answer out loud
Talk to the examiner and it answers back. The model answer is a click away in Feedback and Coaching modes, and never available in Exam Mode.
- 04
Debrief, then repeat the weak part
Your words beside the model answer, every marking point covered and missed, and the priorities to work on next.
Feedback that points at a sentence, not just a score.
Every comment is tied to the words you actually said and the marking point it maps to. The band tells you whether you are covering the ground; the evidence underneath tells you what to change.
- A question bank
- Tests recall in writing. It never hears you run out of structure at minute six.
- A general AI chat
- Will role-play anything, but has no Part B mark scheme behind it and no record of what you keep missing.
- Practising with a colleague
- The best option when you can get it. VoiceOSCE is what you use at 22:00 after a shift, when you cannot.
“I'm sorry that we have not been able to offer you an operation yet. We have had a large number of emergency cases which had to be prioritised. I promise we are doing our best but I cannot give you a date.”
How a station runs
One station takes as long as it takes in the exam. The rest is reading your own words back.
Consent and device check first
Before anything listens you see what is captured, what is stored and how long it is kept. Then a real microphone permission and input-level check — taken once, and only shown again if something changes.
The scenario stays visible
The candidate briefing is read out and written on screen, and stays in the conversation for the whole station.
Three modes, switchable mid-station
Exam holds all feedback to the end. Feedback marks each answer as you go. Coaching offers a hint when you stall. Say “switch to coaching mode” and it changes.
The examiner reacts to what you say
Answers are marked against the station's own scheme as you speak, so the examiner probes what you left out rather than reading from a script.
Debrief against the mark scheme
Marking points covered and missed, the evidence from your transcript behind each one, an OSCE-style band and one clear next action.
What runs the examiner
With a voice model configured, the examiner listens and speaks in realtime and reveals images as it introduces them. Without one, the same station runs on your browser's own speech and a deterministic mark-scheme engine — the questions, model answers and marking are identical either way.
MRCS coverage
Organised against the exam's own blueprint: 17 examined stations across two broad content areas, both of which must be passed in the same sitting. The library shows exactly what is live today.
Anatomy
Applied KnowledgeFacts as they apply to the clinical question, not textbook minutiae.
Surgical Pathology
Applied KnowledgeClinically relevant pathology framed the way the exam asks it.
Surgical Science & Critical Care
Applied KnowledgePhysiology and critical care reasoning held up under questioning.
Clinical & Procedural Skills
Applied SkillsProcedures described start to finish, with the clinical questions that follow.
Communication Skills
Applied SkillsOne skill set, endless contexts — history taking, breaking bad news, consent, the angry patient.
×n = stations that group supplies in the real exam · 17 examined stations, 9 minutes and 20 marks each.
Built on the Pass the MRCS question and explanation base
Written and reviewed by the same clinicians who write the Pass the MRCS material, and published through a draft-and-review flow.
Authorship
Stations are written by the Pass the MRCS clinical team from the same question and explanation base as the main product.
Review
Every station is published through a draft-and-review flow in the content system; nothing reaches candidates unreviewed.
Revision notes
The written notes behind the stations are published alongside them, so the reference and the rehearsal say the same thing.
Corrections
Every station and debrief has a report control. Reported content goes straight to the clinical team's queue.
Feedback is generated against a written mark scheme for each station. It does not assess your clinical judgement, it does not predict whether you will pass, and it is not clinical advice.
Frequently asked questions
Still have a question?
If the answer is not here, send us an email and the clinical team will come back to you.
Send emailNo. Your microphone is used to hear the answer you are giving; audio is transcribed and then discarded, and is never stored for replay. The transcript of what you said is kept with the attempt so the debrief can quote it back to you, and you can delete an attempt at any time.
It answers. Each response is marked against that station's written mark scheme and the examiner replies to what you actually covered — probing what you left out, or moving on when you have it. The station's questions and model answers are fixed and clinically reviewed; the conversation around them is not.
No. Feedback is measured against a written mark scheme for each station. It cannot judge everything a real examiner judges, and it is not a pass prediction. Treat the band as a way to see whether you are covering the ground, not a forecast.
It is held against your browser for the session. If you create an account in the same session, the attempt is kept and moves into your history; otherwise it is discarded.
Yes. The station is one column built for a phone first, with the microphone control within thumb reach and the scenario in the same scroll as the conversation.
You are told which check failed — permission denied, no input device, or an unsupported browser — and offered typed answers so you can still complete the station and the debrief. Voice input needs Chrome, Edge or Safari.
Yes. Attempts, their transcripts and your notes are yours; delete them from your history or delete the whole account from Account settings.
The station bank is being ported from Pass the MRCS and published in batches, organised against the exam's own blueprint — 17 examined stations across Applied Knowledge and Applied Skills. The library shows exactly what is live today.
Rehearse one station now
Unlimited repeats, no account, nothing to install.