AI roleplays that argue back.
Live voice conversations with counterparts who interrupt, stall, remember what you said three turns ago, and change tack when you do. No script anywhere.
Free minutes on sign-up, no card.
The engine
No script. No decision tree.
Say the same unexpected thing to both. The last generation walks a flowchart; this one listens.
“Honestly? Our onboarding is our weak spot.”
→ node 7
→ node 12
“I understand. Our pricing is very competitive.”
Node 7 fires. The tree can’t hear you.
“Appreciate the honesty. So how do you de-risk my first month?”
A listener’s reply — it reacts, it doesn’t retrieve.
The counterpart
Behavior, not answers.
Every scenario casts its counterpart with a behavioral brief — how they stall, what convinces them, when they walk away. Difficulty doesn't change the questions; it changes the person.
Sofia BergmanHead of ProcurementCounterpart brief
EasyCounterpart brief
HardSame person, two editions — pick Easy while you learn the moves, Hard when you want the fight.
The range
One engine, every hard conversation.
The same counterpart engine runs a cold call, a job interview, a layoff, a hostile press interview and a Spanish lesson — each cast and judged by its own definition of good.
25+
scenarios, each with its own counterpart
8
categories, from cold calls to crisis calls
10
languages at native conversational speed
5
rubric dimensions scored per scenario
FAQ
Questions, answered.
How is this different from scripted roleplay tools?
There's no decision tree. The counterpart is a behavioral character, not a script — it responds to what you actually said, follows you when you change subject, and remembers your earlier claims. Two runs of the same scenario never play the same way.
What does a counterpart actually do in a call?
It behaves the way its scenario demands: stalls, objects, pushes on price, probes vague answers, warms only when you earn it — and on hard difficulty, hangs up on a weak opener.
Do I need to write a script or set anything up?
No — pick a scenario and a difficulty and start talking. If you want the counterpart to be a specific real person, paste a LinkedIn URL or website and it's researched into a persona first.
Which scenarios does the engine cover?
25+ scenarios across eight categories — selling, interviews, leadership, negotiation, customer conversations, public & media, language practice and exam prep — each with its own counterpart behavior and its own scoring rubric.
What are the voices like?
Native-quality neural voices at real conversational speed, in ten languages — with the natural hesitations, interruptions and pace of a person on the phone.
What does it cost to try?
Sign-up is free and comes with free practice minutes. No card required.






No script.Just the conversation.
Pick a scenario, pick a difficulty, and say the first word out loud. The counterpart takes it from there.
Start a callPractice makes it perfect — free minutes on sign-up, no card.