> ## Documentation Index
> Fetch the complete documentation index at: https://docs.heymilo.ai/llms.txt
> Use this file to discover all available pages before exploring further.
> Run AI voice (phone) or video (browser) interviews — natural conversation, follow-ups, and scoring.
# Voice/Video Interviews
### When to use it
| Good fit | Skip or sequence earlier |
| ---------------------------------------- | ------------------------------------------------------------------------------ |
| Depth, conversation, rubric-based scores | Yes/no-only volume screens → try **SMS** or **resume** first, then voice/video |
* **Phone**: Easy for candidates. **Video** — Brand + face on camera. Same setup idea.
* **Scored**: Communication and substance, so you can compare fairly.
### Add a Voice Interview stage
The **Voice Interview** card has **three tabs**: **Settings** · **Conversation** · **Tools**. Work through them, then **questions**, then **Save** / **Test**.
#### Step 1 — Add the stage
**Interview Stages** → add **Voice Interview** → open the card.
#### Step 2 — Settings (tab)
| Area | What to set |
| -------------- | ---------------------------------------------------------------------------------------------------------- |
| **Identity** | **Voice** (e.g. Sarah), **Agent Name** (often from Essential Settings) |
| **Modality** | **Voice**, **Video**, or **Phone Calls** (if numbers are set up) |
| **Experience** | **Show Transcript** · **Presence Monitoring** · **Detect suspicious activity** (recommended for web/video) |
#### Step 3 — Detect suspicious activity
Turn on **Detect suspicious activity** to flag signs of cheating during interviews, such as coached answers or a second person in the room, and show them on the candidate report. Flags are for review only — they do not auto-reject candidates or change scores. See [Interview Integrity](../../tools/ai-settings/interview-integrity) for what gets flagged and how to read results on the **Verify** tab.
#### Step 4 — Advanced voice settings (pacing, warnings, speech clarity)
**Where:** Voice/Video interviewer → scroll to the **Advanced** section.
Start with the **defaults** and adjust one field at a time.
What you can customize
* **Visual** — Backdrop, avatar (video/voice look).
* **Timing before AI speaks / after interruption** — Replaces the old **Low / Medium / High** preset with exact values you set.
* **Max reminders** — Maximum number of times the AI can re-prompt a candidate who has gone quiet.
* **Silent seconds before prompt** — How long the AI waits (in seconds) before nudging the candidate.
* **Warn before ending** — Give the candidate a **warning** before the interview ends. Set how many seconds before the end to warn.
* **End interview after N seconds** — How long of silence / inactivity triggers the interview to end.
* **Warning attempts** — How many warnings the AI gives before it actually ends the interview.
* **Minimum speech clarity** — Answers scoring below this threshold are **flagged for review** (informational; helps you spot low-quality audio or unclear responses).
* **Closing Guidance** — default end-of-interview message
* **AI Model** — Conversation model (e.g. **Advanced Model (v3)**).
**Built-in validation**
Some values **must be consistent with each other**. For example, **Warn before ending** must be **shorter** than the **End interview after** time. If the values do not make sense together, the product will flag it and prevent you from saving until you fix it.
**How to tune it**
1. Start with the **default values**.
2. Change **one field at a time**.
3. Run **Test Now** as a candidate and listen for pacing.
**Tip:** Treat this as fine-tuning. For most teams the defaults are a good starting point, and small changes (for example trimming a few seconds off silence tolerance) go a long way.
#### Step 5 — Conversation (tab): structure & messages
* **Questions**: Slider (often **5** is a sweet spot). **Qualification Requirements** — optional pass bar.
* **Messages**: Intro, optional **Welcome Back**, closing. Tokens: `{{name}}`, `{{interviewer_name}}`, `{{company_name}}`, `{{job_title}}`. **Load Defaults** resets copy.
#### Step 6 — Conversation (tab): context
* **Context**: **Enable Context**, **Populate from JD** if you want. Role/company/process fields so questions stay on-brief.
#### Step 7 — Tools (tab)
**Voice Interview** → **Tools**. The tab includes **Tools that extend what the agent can do during the interview** — leave off anything you don’t want live.
* **Resume Questions**: **Résumé-based prompts at the opening** of the call. Put **Resume** (or résumé upload) **before** Voice in the workflow.
* **Speak to Recruiter**: Candidate can **escalate to a human**. **Subscribers** to this agent are notified, including **Slack/Teams** if connected.
* **Candidate Questions**: **Stores what they ask you** during the interview, **one by one**, with a **suggested answer** when HeyMilo can supply one.
#### Step 8 — Questions & evaluation
* **Enforce Objective Evaluation**: ON = must hit a **minimum score** to pass.
* Build the list: **Reorder**, **Re-use**, **+ Add New Question**. Per question: criteria, **Score of 5 / 1**, **Weight**, **Max Follow-ups**. Reuse from Question Library.
#### Step 9 — One question (modal)
**+ Add New Question** → text → **Generate Evaluation Criteria** if you want help. **Scored** rubric or **Structured** pass/fail. Optional: **Don’t include in overall score**.
#### Tags (optional)
**Tags** tab (beside interview questions): extract traits for the report only — **no impact on score**. → [Interview Tags](interview-tags)
#### Step 10 — Test & go live
**Test Now** as a candidate → check audio/video, flow, follow-ups → activate.
**After interviews:** scores, transcript, and cheat detection flags (if on) under the **Verify** tab on the candidate profile. **Résumé-based prompts** on the voice **Scorecard** (**Upfront** / **Deferred** / **Not Asked**) → [Reviewing Candidates](../../getting-started/log-in/reviewing-candidates).
#### Hide scoring criteria from candidates (voice / video)
Some teams do not want candidates to hear or see scoring detail during the interview. HeyMilo can still score answers using your rubric; only the live rubric is hidden.
Where: **Interviewer** → **Questions & Scoring** (voice/video) → Edit the question.
**Steps**
1. Set context and criteria as usual (used for scoring).
2. Leave Don’t include in overall score unchecked so Hide criteria from interviewer appears.
3. Turn on Hide criteria from interviewer.
**Candidate experience**: The interviewer stops surfacing that question’s scoring criteria in the live session (with a short in-product notice). Reports and workflow still use your criteria.
**Follow-ups**: Default follow-ups tied to the hidden rubric may be limited. Use **Interviewer Instructions** under **Advanced Criteria** in the question editor to script follow-ups in your exact wording (including fixed probes that do not depend on the prior answer). Save when done. See [Follow-up questions](/guides/build-your-interviewer/creating-an-interviewer#follow-up-questions) for the full setup.
### Next steps
* [Question Library](question-library) — Reuse questions when creating your agent
* [Interview Tags](interview-tags) — Configure interview tags (Tags tab)
* [Interview Integrity](../../tools/ai-settings/interview-integrity) — Integrity settings and cheat detection
* [Reviewing Candidates](../../getting-started/log-in/reviewing-candidates) — Scorecard, transcript, exports
* [What Candidates See](../../candidate-experience/candidate-experience/what-candidates-see) — Candidate flow