What is a custom AI avatar?
A custom AI avatar, sometimes called a digital twin of a person, is an AI video presenter built from footage of a specific real person, such as a trainer or executive. Once it is created, it can present any new script in that person's likeness without them filming again.
Example
A software company’s head of customer education presents every product update video herself. Each one takes most of a day to film and edit, so the videos lag behind the releases they describe.
She records a short session on her webcam, signs a consent form, and the team creates a custom AI avatar of her along with a clone of her voice. Now the team writes the script for each release, generates the video with her avatar, and sends it to her for approval. Customers still see and hear a presenter they recognize, and she spends her time checking scripts instead of filming.
How it works
A custom AI avatar works like any other AI avatar. The difference is where the likeness comes from: one specific, consenting person instead of a stock actor.
Recording the source footage
The person records footage that the AI model learns from. What is needed depends on the quality level:
- Photo-based: a single clear image. Fast to make, but the face has a limited range of expression.
- Short video: a brief clip of the person speaking to camera, often recorded on a phone or webcam.
- Studio recording: several minutes of footage with controlled lighting and a good microphone, sometimes against a green screen, for the most realistic result.
Vendors usually provide a recording guide that covers what to say, how to frame the shot, and what to avoid, such as busy patterns or hands covering the face.
Building the avatar
The platform uses the footage to model the person’s face, expressions and head movement. Video-based avatars can also pick up the person’s speaking style, such as how they move while they talk. The finished avatar can then be animated to match any new audio.
Adding the voice
The avatar can speak with a stock synthetic voice or with the person’s own voice through voice cloning. A cloned voice is made from a separate audio sample and needs the same care with consent as the face.
Generating videos
From then on, the team writes a script, selects the custom avatar and generates the video. The person does not need to be present, though they should review content published in their name. Because the avatar says whatever is in the script, many organizations require the person, or someone they nominate, to sign off before a video goes out.
Consent and governance
A custom AI avatar reproduces a real person’s face, and often their voice. That makes consent and control the most important part of the process.
- Explicit, informed consent: the person agrees in writing, knowing how the avatar will be used, by whom and for how long. Reputable vendors check for consent before they build an avatar.
- Review rights: agree whether the person approves every script or only certain types of content.
- Access control: limit who in the organization can generate videos with the avatar.
- Exit terms: decide in advance what happens to the avatar and existing videos if the person leaves or withdraws consent.
- Disclosure: tell viewers when a video uses an AI version of a real person, as with any other synthetic media.
Custom AI avatar vs digital twin
People often call a custom avatar a “digital twin”, as in “a digital twin of our CEO”. In engineering and operations, a digital twin means something else: a virtual model of a physical asset, process or system, fed with live data so teams can monitor it or run simulations.
The two share a name but not a technology. A custom AI avatar reproduces how a person looks and sounds on video. It does not copy their knowledge, judgment or decisions, and it only says what someone writes in the script.
Why it matters for L&D teams
- Experts without the filming time. Subject matter experts and leaders can front training while spending their time on review instead of recording.
- A recognizable presenter. Learners see a trainer or leader they know, which can make internal content feel more personal than a stock presenter.
- Updates and translations without reshoots. Edit or translate the script, and the same person presents the new version.
- Clear rules from day one. A written consent and usage policy protects the person whose likeness is used and keeps learners’ trust in the content.
How Colossyan helps
Colossyan offers two kinds of custom avatar. An Instant Avatar is made from a photo or a short phone or webcam video and is ready in under a minute. A Studio Avatar, for Enterprise customers only, is built from 1 to 5 minutes of higher-quality footage. Either can be paired with a clone of the person's own voice. Colossyan only creates an avatar of someone who has given explicit consent, and a custom avatar stays private to your account.
See custom avatars in Colossyan
