> ## Documentation Index
> Fetch the complete documentation index at: https://docs.interviewflowai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Create a role play

> Describe a scenario in the Scenario Studio, review the AI draft, and publish it.

You build role plays in the **Scenario Studio**. You describe the situation in your own words, the AI builds a full draft, and you review and edit it before you publish.

<Note>
  Only Owners can create, edit, and publish role plays. Viewers can open role plays and read results.
</Note>

## Step 1: Describe the role play

1. Click **Role Plays** in the **Main** section of the sidebar.
2. Click **New role play**.
3. In the prompt box, describe the role play in detail.

<Tip>
  Describe the Candidate's role, the AI character, and what success looks like.
</Tip>

The prompt box has a few extra tools:

* **Template**: start from a ready-made scenario. Picking one fills the prompt box with its description, and you can edit it before you send.
* **No job**: choose an Interviewer to use its job description as context.
* **Session settings**: set the language, webcam, cold call, and timing before the draft is built. You can change these later.
* **Clear**, undo, and redo.

Available templates include:

| Template | Difficulty |
| - | - |
| **SaaS discovery call** | Medium |
| **Cold call (SDR)** | Medium |
| **Pricing objection** | Hard |
| **Frustrated customer escalation** | Medium |
| **Production incident debugging** | Medium |
| **Renewal at risk** | Medium |
| **Difficult feedback** | Medium |

Press Enter or click the send button.

## Step 2: Answer the assistant's questions

The assistant may ask clarifying questions. Type your answer, or click **Just build it** to skip the questions and build the draft straight away.

While it builds, you see its progress: writing the scenario, creating the character, and writing the evaluation criteria.

## Step 3: Review the draft

The draft opens in the editor as a **Draft**. The editor has two panels:

* **Chat** on the left: ask the assistant to make changes.
* Four tabs on the right: **Basic details**, **Scenario details**, **Character details**, and **Session settings**.

On a phone, use the **Chat** and **Scenario** switch at the top to move between the panels.

### Use the chat assistant

Type a request such as "make the character more sceptical" or "add a criterion about closing". The assistant updates the form and lists what it changed, marked **Updated · not saved yet**.

<Warning>
  Changes from the chat assistant are not saved automatically. Review them in the tabs, then click **Save**.
</Warning>

### Save your work

The editor doesn't autosave. The header shows **Unsaved changes** when you have edits.

* Click **Save** to save the draft.
* To undo everything since your last save, click the **More actions** menu above the chat and choose **Discard unsaved changes**.
* If you try to leave with unsaved changes, you are asked to confirm.

## Basic details

| Field | What it does |
| - | - |
| **Title** | The role play's name. Required to publish. |
| **Scenario description** | A short summary shown on the role play card and page. The Candidate doesn't see it. Required to publish. |
| **Introduction (optional)** | Spoken before the character's opening line. It explains the Candidate's role and the context. You can use placeholders such as `{firstName}`, `{personaName}`, and `{durationMinutes}`. Click **Reset to default** to restore the standard text. |
| **Difficulty** | **Easy**, **Medium**, or **Hard**. See below. |
| **Category** | Such as **Sales discovery**, **Cold call**, **Objection handling**, **Customer support**, **Technical debugging**, **Account management**, **Negotiation**, **People management**, or **Custom**. |

### Difficulty

Difficulty sets the character's default personality and how strictly the transcript is graded.

| Difficulty | Character | Grading |
| - | - | - |
| **Easy** | Cooperative and shares facts readily. | Gives credit for reasonable attempts. |
| **Medium** | Neutral, with standard sharing. | Expects a competent professional response. |
| **Hard** | Guarded and needs deeper questions. | Expects precise, complete answers. |

## Scenario details

This tab holds what the Candidate knows and how they are graded.

* **Scenario summary**: **Your role** and **Situation**. The Candidate sees these on the join page and during the call. Both are required to publish.
* **Scenario objective**: what the Candidate is trying to achieve. Required to publish.
* **Evaluation criteria**: what the Candidate is scored on. See below.
* **Candidate visibility**: what the Candidate sees of the criteria. Choose **Criterion names only (recommended)**, **Names and what we look for**, or **Hidden**. Strong and weak answers are never shown.
* **Candidate materials**: up to five plain-text items, such as a price sheet or product facts. Click **Add material** and enter a **Title** and **Content**. The Candidate sees them on the join page and beside the call.
* **Tips for the candidate (optional)**: short advice shown to the Candidate.

### Evaluation criteria

You need at least three criteria to publish. You can add up to 12.

Click **Add criteria** to open the criterion editor:

* **Name**: what the criterion measures. Required.
* **Looking for**: the evidence you want in the Candidate's answers.
* **Strong answer (5)**: what full marks look like, as observable behaviour.
* **Weak answer (1)**: what a score of 1 looks like.
* **Weight**: from 1 to 10. The default is 5. A weight of 1 is informational: it appears in the report but isn't scored.
* **Must pass**: a score of 2 or lower is flagged in the report and makes the AI recommendation **No**. It doesn't change the score.
* **Skill**: an optional tag, such as Discovery or Objection handling.

Click **Save changes** in the editor. Name, looking for, strong answer, and weak answer are all required to publish.

Other actions:

* Use the menu on a criterion to **Edit**, **Duplicate**, **Move up**, **Move down**, or **Delete** it.
* Click **AI-tune** to rewrite the strong and weak answers as observable behaviour.
* Click **Clear** to remove all criteria.

## Character details

This tab sets up the AI character. Sections marked **Hidden from candidate** are never shown to the Candidate.

### Basic identity

* **Photo**: click **Choose avatar** to pick a photo avatar or a live avatar. Photo avatars are still headshots. Live avatars are animated characters that speak on the call. You can also click **Upload photo** to use your own square image, at least 150×150px and up to 1 MB. Click **Remove** to clear it.
* **Name**, **Job title**, and **Company**: all required to publish.
* **Voice**: the voice the character speaks with.

### Conversation

* **Opening line**: spoken word for word when the call starts. Required to publish.
* **Backstory**: what the character knows about themselves, their company, and their situation. Required to publish.

### Situational understanding (hidden facts)

Hidden facts are details the character reveals only when the Candidate earns them. They test how well the Candidate asks questions. You need at least one hidden fact to publish, and you can add up to 15.

Click **Add fact**, then fill in:

* **Fact**: the detail, such as "Budget is \$40k, approved in Q1".
* **Shares it**: when the character reveals it:
  * **Freely**: can come up naturally.
  * **If asked**: shared when the Candidate asks a relevant question. This is the default.
  * **If probed deeply**: shared only after a follow-up that builds on an earlier answer.
  * **Never**: context for the character only.
* **Trigger (optional)**: what the Candidate must say or ask before the character shares it.

**Ground truth (optional)**: for debugging or support scenarios, enter the real cause and fix. The character stays consistent with it but never states it.

### Personality

Choose **Traits** to set six sliders from 1 to 5: **Assertiveness**, **Patience**, **Talkativeness**, **Technical depth**, **Warmth**, and **Scepticism**. The defaults come from the difficulty. Click **Reset to medium defaults** (or easy or hard, to match the difficulty) to restore them.

Choose **Custom** to describe the personality in your own words instead.

### Conversation guidelines

Add rules for how the character responds in specific situations. Click **Add text section** for a plain rule, or **Add triggers section** for situation-based responses. Each trigger has **When the candidate says or does**, one or more responses (click **Add response** to add more, up to four), and an optional **Resolved when**. Click **Clear all** to remove every section.

Use the responses to describe how the character should react if the Candidate leaves a concern unresolved. Review them alongside the difficulty setting.

### Success and ending

* **Success condition**: when the objective counts as achieved. Required to publish.
* **Concession rule (optional)**: when the character should become more open.
* **End conditions**: when the call should end. Four are included by default, such as reaching the time limit. Keep at least one.
* **End-call phrases (optional)**: up to five phrases the character can use to end the call.

## Session settings

| Setting | What it does | Default |
| - | - | - |
| **Scenario language** | The language for the AI character and the introduction. | English |
| **Enable webcam** | The Candidate appears on video. | Off |
| **Desktop only** | The Candidate must join from a laptop or desktop computer. | Off |
| **Enable cold call settings** | Simulates an incoming call with a ring tone. The character may hang up if the Candidate gives no reason to stay. Set **Ring time** (2–6 seconds) and **Hang up if no value within** (20–90 seconds, or leave empty for never). | Off |
| **Duration** | How long the call lasts, from 3 to 15 minutes. | 8 minutes |
| **Link expiry** | How long an invite link stays valid, from 1 to 30 days. | 4 days |

## Publish the role play

Click **Publish**. If anything required is missing, the tabs with missing fields are marked with a dot, and the editor opens the first one.

Once published, the role play's status changes to **Published** and you can invite Candidates. See [Invite Candidates and review results](/role-plays/invite-and-review).

## Edit a published role play

1. Open the role play from the **Role Plays** page.
2. Click **Edit**, or click the pencil on any section of the **Overview** tab.
3. Make your changes.
4. Click **Save changes**.

If Candidates have invites they haven't started yet, you are asked to confirm with **Change the live role play?**

## Clone, archive, or delete

Open the role play and click the **More actions** menu:

* **Clone role play**: makes a copy and opens it in the editor.
* **Archive**: marks the role play as archived.
* **Delete role play**: removes it from your role plays. You can't delete a role play that has invites that haven't finished.

## Related docs

* [Role plays overview](/role-plays/overview)
* [Invite Candidates and review results](/role-plays/invite-and-review)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.