Docs
Training a Custom Avatar

Training a Custom Avatar

Turn a single photo or a short video of yourself into a reusable AI avatar. This guide covers both training paths — photo avatars and digital twins — including the consent step and what each status means.

VideoMule can train an AI avatar of you (or anyone on your team) and use it to present your videos. Training happens in the Avatar Library — open it from the dashboard sidebar and scroll to the Training ground section.

There are two ways to train:

  • Photo avatar — trained from one clear photo. Ready in a couple of minutes.
  • Digital twin — trained from real footage of you speaking. Higher fidelity, but it requires a short identity consent recording before it activates.

Option A: Train a photo avatar

  1. In the Training ground, find the Photo avatar card.
  2. Click Upload a photo and pick a JPG, PNG, or WebP file up to 10 MB.
  3. Enter an Avatar name (for example, My Presenter).
  4. Click Train photo avatar.

Photo requirements:

  • One person, looking at the camera
  • Well-lit face — no sunglasses, hats, or heavy filters
  • Close-up or waist-up, business-like framing

The avatar shows a Training badge while it processes, then switches to Ready — usually within a couple of minutes. Photo avatars don't need a consent recording.

Option B: Train a digital twin

  1. In the Training ground, find the Digital twin card.
  2. Click Upload footage and pick an MP4, WebM, or MOV file. The footage must be between 15 seconds and 10 minutes long and no larger than 200 MB.
  3. Enter an Avatar name.
  4. Click Train digital twin.

Footage tips for the best results:

  • Talk directly to the camera for the whole clip.
  • Use stable framing, good lighting, and clean audio.
  • Keep just you in frame — no cuts, overlays, or background music.

After a digital twin finishes its first training pass, it shows an Awaiting consent badge. Before the twin can be used, the person in the footage must record a short identity consent statement:

  1. On the avatar's card in the library, click Record consent.
  2. A secure, hosted webcam page opens. Read the short consent statement aloud and submit the recording.
  3. You are redirected back to the Avatar Library automatically.

Once the consent is approved, the twin finishes training and switches to Ready. Status updates aren't instant — the library checks automatically when you open the page, and you can click Refresh status at any time.

Avatar statuses

BadgeMeaning
TrainingThe avatar is being trained. Photo avatars take minutes; twins take longer.
Awaiting consentDigital twins only — the identity consent recording is still needed.
ReadyThe avatar can be selected and used in your videos.
FailedTraining didn't succeed. Failed avatars don't use up a slot — delete the failed attempt and try again with a better photo or footage.

Adding more looks

An avatar isn't limited to a single appearance. Open your avatar in the library and click Add look to upload another photo (or more footage for a twin) of the same person — a different outfit, pose, or setting.

  • Extra looks don't use a new avatar slot.
  • Twins don't need a new consent recording for additional looks.
  • Each avatar can have up to 10 looks.

Using your avatar

Once a look is Ready, it appears alongside the stock avatars when you choose an avatar for a project. See Placing Avatars in Your Video for how avatars are placed as intros, outros, and transitions.


Frequently Asked Questions

No. Only digital twins (trained from real footage) require the identity consent statement.

Does deleting an avatar free up a slot?

Yes. Deleting an avatar (the person and all of their looks) releases the slot. Deleting a single look keeps the slot occupied as long as other looks of that person remain.

My footage upload was rejected. Why?

Check the constraints: MP4, WebM, or MOV format, between 15 seconds and 10 minutes long, and at most 200 MB. Trimming or compressing the file usually fixes it.

Can someone else on my team be an avatar?

Yes — but for digital twins, the person in the footage has to record the consent statement themselves. Send them the consent link when the avatar reaches Awaiting consent.