> For the complete documentation index, see [llms.txt](https://herd-security.gitbook.io/herd-security-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://herd-security.gitbook.io/herd-security-docs/trainings/getting-started-with-trainings/custom-likenesses.md).

# Custom Avatars and Voices (Executive Deepfake Videos)

Deepfake your own executives, with their consent: build an AI avatar and cloned voice of a real person from one clip, recorded in the browser or uploaded, and use it in your organization's training vi

## Overview

Talking-head AI videos normally use an avatar and voice from the HeyGen library. A **custom likeness** lets you use a real person from your organization instead — your CEO or CISO welcoming new hires, a team lead introducing a policy — built from a single clip of them speaking to camera. The clip's video becomes the avatar and its audio becomes the voice, and the likeness can then say any script you write.

This is how you make a **video deepfake of an executive** for training: an executive records themselves in the browser, consents, and once the likeness is ready their AI double can deliver a training, for example one that shows staff how convincing a deepfaked leader can be and why urgent requests need checking through a second channel. It's a video creation tool for trainings, not a simulation: Herd never uses a likeness in a phishing, voice or other simulation. For the hands-on voice-cloning exercise, see [Deepfake Phishing Simulations](/herd-security-docs/simulations/deepfake-phishing-simulations.md).

Likenesses are opt-in for the person concerned and stay inside your organization:

* The person is emailed for **consent**, and the likeness can't be used until they give it.
* It is used only in **your organization's training videos** — never in simulations, and never by another organization.
* The person can **withdraw** at any time, which deletes the likeness and takes down every video made with it.

{% hint style="info" %}
Creating and managing likenesses needs the **Org Settings** permission (admins always have it), on top of talking-head video being enabled for your workspace. See [Create Training with AI](/herd-security-docs/trainings/getting-started-with-trainings/create-training-with-ai.md) for talking-head videos in general.
{% endhint %}

***

## Creating a likeness

1. Open **Generate AI video** — from a training step's **Attach Content → Generate AI video**, or from the [Media Library](/herd-security-docs/trainings/getting-started-with-trainings/media-library.md)'s **Generate media → AI video**. Both open the same screen.
2. Choose the **Talking head** style. In the **Avatar and voice** section, click **Create custom avatar and voice**.
3. The **Custom avatars and voices** dialog explains what gets made and lists your organization's existing likenesses and their status. Fill in the form to add one:
   * **Person** — search your directory by name or email. This is who will be emailed for consent. Making one of yourself? Click **It's me**.
   * **Name in the picker** — how the likeness is labelled in the Avatar and Voice pickers.
   * **Video clip** — choose **Record with your camera** or **Upload a video** (browsers that can't record show the upload only). The clip's audio becomes the voice.
4. Click **Create and email for consent** (or **Create and email me the consent link** when the likeness is of you). Herd uploads the clip and emails the person; the toast confirms *We've emailed \<name> to ask for their consent.*, or *Check your inbox* for yourself.

Nothing is built until the person agrees.

### Recording in the browser

With **Record with your camera**, click **Turn on camera**, then **Start recording**. The stop button reads *Keep going: Ns to the minimum* until you pass 15 seconds, then **Stop recording**; recording stops on its own at 10 minutes. Watch the take back under **Your recording**, then keep it, **Record again**, or **Discard take**.

For a good likeness:

* Face the camera in even light, with only you in frame.
* Talk naturally, about anything, for one to two minutes. More speech makes a better voice.
* Keep a quiet room and pause now and then with your mouth closed.

### Uploading a clip

With **Upload a video**, pick an **MP4, MOV or WebM** file, **15 seconds to 10 minutes long and up to 200 MB**, of the person speaking to camera. Use this to make a likeness of someone else from footage they've recorded.

***

## What the person receives

The person gets an email titled ***\<your name> asked to create an AI likeness of you***, headed **Your consent is needed**. This is true even when you made the likeness of yourself: consent is always recorded from the email, and the dialog tells you to check your inbox. It explains that the likeness is used only in your organization's training videos, never in simulations, and carries two links:

* **Review and record consent** — opens HeyGen's consent page, where the person records a short consent statement. This link **expires in 24 hours**; an admin can send a fresh one with **Resend link** while the likeness is still waiting.
* A **withdraw consent and delete the likeness** link, which **never expires**, so the person can revoke later without asking anyone (see [Withdrawing consent](#withdrawing-consent-and-deleting-a-likeness)).

***

## Likeness status

Each likeness in the **Custom avatars and voices** dialog shows one of:

| Status                  | Meaning                                                                                                                                                                                         |
| ----------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Waiting for consent** | Emailed to the person for consent. It can't be used until they agree.                                                                                                                           |
| **Training**            | The person has consented and the avatar and voice are being built. This can take a few hours. When it finishes, the admin who created it gets one email, ***The "\<name>" likeness is ready***. |
| **Ready**               | Ready to use in this organization's talking-head videos.                                                                                                                                        |
| **Failed**              | Setup failed. Delete it and try again with another clip.                                                                                                                                        |
| **Revoked**             | Consent was withdrawn or an admin deleted it. Its avatar, voice, and videos were deleted.                                                                                                       |

***

## Using a likeness in a video

Once a likeness is **Ready**, the **Avatar** and **Voice** pickers on the Generate AI video screen gain a **Your organization** group, listed above the **HeyGen library**. Pick the person's avatar and voice like any other and generate as usual; the rest of the screen (script, filters, previews, remembered choices) works the same way. Only admins in your organization can see or use these entries.

To set both at once, open **Create custom avatar and voice** from the Generate AI video screen and click **Use in this video** on a **Ready** likeness. Herd sets it as the video's avatar and voice and closes the dialog.

### Reviewing the video before learners see it

A talking-head video generated for a training step isn't put on the step when it finishes. Watch the take first: **Use this video** adds it to the step, and **Generate another** replaces it without it ever reaching the step. Closing the page also adds the latest take. Every take is still saved to the [Media Library](/herd-security-docs/trainings/getting-started-with-trainings/media-library.md).

***

## Withdrawing consent and deleting a likeness

Either side can end it, and the effect is the same:

* **The person** opens the withdraw link from their consent email, which lands on a page headed **Withdraw consent and delete your likeness** (no Herd sign-in needed). It says what will happen, and **Withdraw consent and delete** carries it out; the page then confirms **Consent withdrawn**. A link that has already been used shows **Already withdrawn**.
* **An admin** clicks **Delete** on the likeness in the **Custom avatars and voices** dialog and confirms in **Delete this likeness?** with **Delete and take down videos**.

In both cases Herd deletes the avatar and voice at the provider and **removes every video made with them** from the training steps and Media Library items that used them. It can't be undone. The admin who created the likeness is emailed (***The "\<name>" likeness was revoked***) with the list of trainings that were affected, so they can re-record those steps with a different avatar.

{% hint style="warning" %}
A step whose video is taken down is left without its media. Check the trainings named in the revocation email before they next go out.
{% endhint %}

***

## Related

* [Create Training with AI](/herd-security-docs/trainings/getting-started-with-trainings/create-training-with-ai.md) — the AI video dialog, avatar and voice pickers, and previews.
* [Media Library](/herd-security-docs/trainings/getting-started-with-trainings/media-library.md) — where generated videos are stored.