RoleFit Help

On the conversation screen, press 'Start conversation' or 'Start practice' and your microphone turns on. If your browser asks for microphone access, choose 'Allow'. When 'Please speak now' appears, just talk; don't press the microphone button. If the avatar greets you first, wait until the greeting ends.

No. In the default mode, Auto Send, your microphone is already on during your turn. Pressing the button switches it off, and it stays off until you press it again. If that happens, press it once more to switch it back on.

No. Only one person speaks at a time, so anything you say while the avatar is speaking or preparing its reply is not sent. Your browser's microphone indicator stays on for the whole conversation, but your words are not recognised at those moments. Speak when the avatar has finished and 'Please speak now' appears. If you use Push to Send, the microphone button is locked while the avatar is speaking or preparing its reply; press it to speak once the avatar has finished.

When you stop speaking, what you have said so far is sent. Say everything you want to say in one go, then wait for the avatar to reply.

With Push to Send, press the microphone button at each of your turns, then speak. When you stop speaking, your words are sent and the microphone switches off again. While the avatar is speaking or preparing its reply the button is locked; once the avatar has finished, press it to say your next line.

They are not required, but the avatar is easier to hear with earphones. If you use a screen reader, always use earphones: its speech can reach your microphone and be sent as your answer.

If you say nothing for 60 seconds during your turn, the conversation ends. A warning appears after 45 seconds. While help or the menu is open, this count stops and starts over when you close it; the practice time keeps running. The conversation also ends when the practice time is up; if the avatar is speaking, it ends after that line. The avatar can end the conversation too: when you wrap up with a goodbye, it says goodbye and ends it. Except at the Easy level, if you keep brushing it off or blaming it, it warns you once and then leaves.

When you finish, the result page shows your turns, what the practice looks at and the AI feedback. When the practice shows the transcript, the result page shows it too. AI feedback needs at least one answer from you. If you practised through a public link, you can reopen the result only in the same browser tab. Results of assigned training stay in Learning history.

No. Using help or AI hints is not recorded in your result. AI hints appear only when the person who built the practice allows them.

In the site settings at the left of the address bar, set Microphone to 'Allow'. Also check that no other tab or app, such as a video call, is using the microphone. If it won't connect on a company network or VPN, try another network.

We recommend the latest Chrome or Edge on a computer. Safari and phone browsers also work, but the microphone or the screen may behave differently on some devices. If you opened the link inside an app such as KakaoTalk, open it again in Chrome or Safari.

For now you can practise only by speaking; there is no way to answer in text. When the practice shows the transcript, you can also read what the avatar says. If you need another way, tell the person who shared the practice.

What you say is converted to text and stored with the AI feedback, and the people running the practice can see it. The transcript and the AI feedback are deleted one year after the practice ends. See our Privacy Policy (opens in a new window) for details.

Start a simulation from a template on Home, or from scratch. Fill in the five builder steps, try it yourself with 'Test simulation' and refine what you need. When it is ready, publish it and share it with learners through a public link or member training (private).

The trainee role, avatar role, situation and training goal are shown to learners as written. Write the goal and the success condition as something you can observe in the conversation, such as 'agrees on a date for the next meeting'. Facts the learner should discover by talking go in Hidden information, not in the situation.

Choosing an avatar sets its face and voice together. Personality, conversation style and how it responds are set in this step and are not shown to learners.

Upload files the avatar may rely on in the conversation, or add them from the Knowledge library. You can publish without any. A file that is still processing must finish first, and a file that failed must be retried or deleted from the Knowledge library before you can publish.

Set the scoring criteria the AI review uses. Three to five criteria keep the result easy to read. With 'Calculate an overall score' on, the weights must total 100% before you can publish. Write 'Must do' and 'Must not' one per line; they are used only for scoring, and the avatar does not know them.

Choose a session length of 5, 10 or 20 minutes, or Unlimited; Unlimited still stops at 20 minutes per session. Each start holds this much training time, and only the time actually used is deducted at the end. With 'Avatar speaks first' on, the avatar opens the conversation with a greeting; with the 'Fixed line' greeting type, it uses the first greeting you wrote.

A public link lets anyone who has it practise without signing in; use it for people outside, such as customers or applicants. Member training creates no link: once you assign the training and open it, only the invited members you assigned can practise. You can use both, and either way the practice uses this workspace's training time.

Publish as member training (private) and invite members. When a member accepts the invitation and joins, you can assign the training to them. Once you assign the training and open it, only the invited members you assigned can practise. An invitation link works for one email address and expires after 7 days. RoleFit does not email it, so send it yourself; if it is lost, create a new link from the invitations list. Members see the training only after you open it, and an opened training cannot go back to 'Not started'.

Builder tests, public-link practice and member training all use this workspace's training time. Each start holds training time equal to the session length and only the time actually used is deducted, so a practice cannot start when less time is left than its length. Public links have no separate share of that time: whatever is left can be used for any practice. With an Enterprise contract that has no time limit, no training time is deducted.

Without a paid plan, a workspace gets 10 minutes a month, with up to 1 person practising at a time. It refills at midnight Korea time on the first day of each month, and unused time does not carry over.

Conversations you test in the builder are recorded on the Sessions page. Open the session and choose 'Review this session' to get the AI review.

If you edit and publish again as a public link, the new version gets a new link. The old link keeps opening the previous version until you switch it off, and a link that is switched off cannot be switched back on. Member training uses the version chosen when the assignment was created.

Choosing 'End assignment' on the training details stops members from starting new practice, and any practice in progress ends without completion. To remove one person, use 'Remove from training' for that member. Neither can be undone, and the records so far are kept.

For now, one workspace owner builds and manages the simulations. Members you invite take part as learners.