Youteach.Opsinisrunstheroom.
It voices the patient, plays the nurse and drives the monitor, so one facilitator can run a full acute scenario and debrief from a timed record of every decision.
Every number on the monitor comes from the case, not the AI.
Every number on the monitor comes from the case, not the AI.
One acutely unwell patient. Everything in the room responds.
Opsinis plays the parts of a scenario that usually take extra hands, so the tutor is free to watch and teach.
The patient
Speaks in his own voice, with a history that fits the case. As he deteriorates he gets more breathless, then muddled.
The nurse
Hears spoken orders and carries them out: oxygen, a cannula, fluids, bloods. A dose out of range, she queries.
The monitor
Follows a physiology model written for the case. Treat him and the numbers respond. Leave him and they drift.
The record
Every question, examination and treatment is timed as it happens, ready for the debrief.
Assessed the way you teach it: A to E.
Every check and treatment has its place in the A–E approach, on the tutor’s tablet and in the debrief. The student leads. The simulation answers.
Is he talking? He answers in short sentences between breaths, so his airway is patent. Like every finding, it’s only there if the student asks for it.
Rate, saturations, and each lung on its own. The chest sounds are real recordings played at the bedside, and what they mean is left to the student.
A pressure, capillary refill, a cannula. Bloods, cultures and a gas come back when they would on the ward, and the fluid they give moves the numbers on the monitor.
Conscious level and a glucose. If his pressure keeps falling he becomes muddled, and the student will hear it in his voice before they see it on a chart.
Temperature and skin, and a look for the source. Then escalate: the registrar, outreach, or a peri-arrest call if he’s that unwell. Each is timed for the debrief.
Is he talking? He answers in short sentences between breaths, so his airway is patent. Like every finding, it’s only there if the student asks for it.
Orders, spoken the way they would be on the ward.
She works out which actions were asked for, and nothing else. Every effect comes from the case, exactly as if the tutor had tapped it. Pick a line to see what she does with it.
“Cannula in, bloods and cultures, and five hundred of Hartmann’s stat.”
- 1IV cannulaDone
- 2BloodsDone
- 3Blood culturesDone
- 4Hartmann’s 500 mlDone
Carried out at once, cannula first. Each is timed for the debrief as a spoken order.
Cannula’s in. Bloods and cultures sent, 500 of Hartmann’s running.
You stay in charge of the room.
Opsinis runs the scenario. You decide how it unfolds, from a tablet, at any moment.
The tutor taps every action
Decide how much the nurse does
Let her act on clear spoken orders, approve each one with a tap, or run the room yourself.
Hold a result back for teaching
A result can wait with you until you’re ready to discuss it. The nurse won’t read it out before then.
Push the scenario when you choose
Nudge a vital sign or trigger a deterioration, within the limits the case allows. The debrief never blames the student for it.
“You’re the FY1 in the emergency department. The nurse has asked you to see David Hughes, 65, who came in this morning with three days of cough and fever…”
Brief them, then begin
Each case opens with a short brief: who they are, where, and who called them. Read it out, show it, or let the nurse read it.
A debrief that’s ready before the room is.
The timings come from the simulation, not from someone’s notes, and every point cites the moment it happened.
- OxygenOn time
- IV accessOn time
- IV fluidsOn time
- LactateNot done
- EscalationLate
- IV antibioticsOn time
- SPatient and problem
- BBackground and treatment
- ANo observations given
- RAsked for a review
- 00:44Oxygen went on first, for saturations of 91% on air.
- 01:36Blood cultures were taken before the antibiotics, as they should be.
- —No venous gas, so his lactate was never seen. It would have shown how unwell he was.
- 09:06The handover gave the problem and the ask, but no observations for the registrar.
Facts from the simulation
Times and findings are recorded by the simulation itself, so the feedback rests on what actually happened.
Spoken and tapped, shown apart
See which orders the student gave out loud, and which the tutor tapped on their behalf.
Fair to the student
A deterioration the tutor triggered is shown as the tutor’s, never as the student’s failing.
The AI does the talking.
Never the medicine.
If the numbers on the monitor can’t be trusted, the simulation can’t either. So no clinical decision is ever left to a language model.
The monitor is never guessed
Vital signs, results and outcomes come from a physiology model written for each case. The language model can’t set a number, and the patient’s words are checked against the case.
Real sounds, real images
Chest sounds and X-rays are real, anonymised clinical recordings, licensed for use. Nothing clinical is AI-generated, and interpreting it is left to the student.
The tutor’s hand is always shown
Anything the tutor changes is logged as theirs, so the debrief never mistakes it for something the student did.
No case is final until a clinician signs it off.
Each clinical value, from the starting observations to how quickly fluids work, is written down for a clinician to check, so the people who teach it can see exactly what the simulation will do.
Every clinical value in plain English, with its unit and its reasoning.
A clinician checks each one and answers the questions it raises.
Named and dated. Until then, the case stays marked as a draft.
The emergencies junior doctors meet on call.
Each scenario can be played with different patients: another age, another background, the same illness.
Sepsis from pneumonia
Three days of cough and fever, and he’s getting worse. Recognise sepsis, start treatment in time, and hand over to the registrar.
Also in the library
In clinical review- Anaphylaxis02
- Acute severe asthma03
- Diabetic ketoacidosis04
- Hypoglycaemia05
- Upper GI bleed with shock06
- High-risk pulmonary embolism07
- Opioid toxicity after surgery08
- Acute pulmonary oedema09
Built to sit beside your sim room.
Opsinis is designed to work alongside the manikin and equipment you already use. Every centre is set up differently, so we shape the setup around yours.
Asked by clinical leads.
Anything else? Ask us during the demo.
Does it replace the tutor?
No. It takes on the patient’s voice, the nurse and the note-taking, so the tutor can watch, teach and decide how the scenario unfolds.
Can the AI get the medicine wrong?
The AI only does the talking. Every vital sign, result and outcome is calculated from the case, and what the patient says is checked against it. The nurse only carries out actions the case allows, and queries doses outside its range.
Are the chest sounds and X-rays real?
Yes. They’re real, anonymised clinical recordings, licensed for use, never AI-generated. Students listen to each lung on its own and interpret the findings themselves.
How long does a scenario take?
About ten minutes, so it fits a normal teaching session with time to debrief. Near the end the nurse puts the registrar on the phone for the student’s SBAR handover.
Which scenarios are ready?
Adult sepsis is the lead scenario, with three different patients. Eight more acute emergencies are written and in clinical review.
Do we need a new manikin?
No. Opsinis is designed to sit beside the manikin you already use. It doesn’t connect to it or replace it.
See a scenario run with your faculty.
We’ll run the sepsis scenario live. Hold the tutor’s tablet, talk to the patient, and give the nurse your own orders.
- A live run of a full scenario
- Try the tutor’s controls yourself
- Questions answered by the team that built it