How to Make an AI Digital Twin of Yourself for Video

On VdoBloom there are two different things people mean by an AI digital twin, and picking the right one first saves you a wasted render. The Spokesperson tool creates a consistent AI presenter — a face the platform generates for you, saved once and reused on every video. If you want a twin that looks like you, you need Avatar Presenter, which drives a photo you upload with a script you write. Neither one is a live avatar that answers questions in real time.

Updated October 2026.

Two meanings, two tools

The phrase “AI digital twin” gets used for a few different products. On this platform it covers two jobs that look similar in a feed and are built completely differently:

• A consistent brand persona. You do not appear. A face is generated for your brand and stays identical across every video, so your channel has a recognisable presenter without you filming anything.
• A stand-in that looks like you. You upload a photo of yourself and the tool makes that photo speak your script. The result carries your likeness, which is why the consent rules around it are stricter.

Both are driven by the same idea — a still image plus a script becomes a talking video — but only the second one is a twin in the literal sense. The Spokesperson tool never asks for a photo of you; it asks you to describe the presenter you want and generates it.

What Spokesperson actually builds

Spokesperson is a two-tab tool. The first tab creates the presenter, the second puts it to work.

On the create tab you describe a face rather than supply one. The inputs are gender (Young Woman, Young Man, Middle-aged Woman, Middle-aged Man), age band (20s through 50s), an ethnicity field you type yourself, a style (Professional, Casual, Medical, Fitness), an optional name, and a free-text details box that starts with a default describing natural skin, a warm and trustworthy look, and business casual attire. The image is generated by the platform’s image model at a portrait crop.

The part that makes it a persona rather than a one-off picture is that the result is saved. It is stored in your browser under a brand-spokesperson key, so the same face comes back on your next visit, and it can also be pushed into My Library as a reusable character. That is the whole value: consistency. A viewer who has seen three of your videos recognises the same presenter in the fourth, which is a thing one-off AI generations never manage.

You can also pull an existing character from My Library instead of generating a new face, which is useful if you already have a consistent character from image work.

Build a saved spokesperson, step by step

1. Open Spokesperson at /dashboard/video-creation/spokesperson/ and stay on the create tab.
2. Set gender, age and style. Start here — these three do most of the work. Medical and Fitness styles read very differently from Professional.
3. Add details in your own words. The details box is free text, so what you write here is the difference between a generic stock presenter and one that fits your brand. Say what you want to see, not what you want to avoid.
4. Generate the face. The default image is a portrait crop, which is what a talking-head video wants.
5. Save it. Give it a name and confirm it is stored. This is the step that turns a generation into a persona — skip it and you start over next time.
6. Switch to the generate tab and move on to the script.

Give it a script and a voice

The generate tab takes a script in a text box, an accent, and a tone. Accent starts at American and is a free-text field, so you can type British, Australian or anything else the model recognises. Tone is a set of buttons — Confident, Friendly, Expert, Energetic, Calm — and starts on Confident.

Then you pick the video model, duration, quality and aspect ratio. The model list here is the image-to-video set, which is the right family: the tool is animating a still face into speech, not generating footage from nothing. The default is ByteDance Seedance 2, which supports 4 to 15 second clips at up to 4K with a choice of six aspect ratios. One model in the list, Seedance 1.5 Pro, generates audio along with the video, and the tool switches on audio automatically when you select it.

Practical point: keep the script short and spoken. These are 4 to 15 second clips. A script that runs two minutes will either be split or truncated, and a truncated presenter video ends mid-sentence.

If you want a twin that looks like you

For that, leave Spokesperson and open Avatar Presenter. It works from three things: a lesson-style script, a presenter photo, and a consent confirmation.

• The photo should be a clear, front-facing shot. This is the whole likeness input, so lighting and sharpness matter more than anything else.
• The script can be long — up to twenty thousand characters, which is roughly a course module — and the tool splits long scripts into numbered parts so each part becomes its own video.
• The framing is one of three: at a desk, seated and waist-up; a plain studio with a clean background, waist-up; or keep my photo, which leaves your own setting behind you.
• The model is Kling Avatar Standard at 720p, Kling Avatar Pro at 1080p, or OmniHuman 1.5, which is capped at sixty seconds of audio. The Kling models handle up to five minutes.
• The voice comes from a list of around twenty stock voices with a default already selected. This is not voice cloning — if you want the video in your own voice, that feature lives inside Creator Reel and not here.

Avatar Presenter also requires the consent checkbox before it will generate. That is not decoration; it is the gate that keeps a photo of someone else from becoming a talking video. If you are not the person in the photo, you need their permission — and for a real person, permission in writing is the only version worth having. The wider question of when an avatar video is the right format at all is covered in AI avatar ads vs UGC creator ads.

There is a third route if you already have audio rather than a script: the avatar tab takes a photo plus an audio file and syncs the face to it, with seven models to choose from. The tool page for that route lists the models and their caps at AI talking avatar.

What it costs

Talking video is priced per second of finished footage on the avatar models, and per clip on the video models.

Route	How it is priced
Spokesperson on Seedance 2 Fast	About 10 credits per second at 720p
Spokesperson on Seedance 2	About 16.5 credits per second at 720p
Spokesperson on Seedance 1.5 Pro with audio	23 credits for an 8-second 720p clip; 48 credits at 1080p
Avatar Presenter on Kling Avatar Standard	About 3.2 credits per second of audio — 96 credits for 30 seconds
Avatar Presenter on Kling Avatar Pro	About 6.4 credits per second of audio — 191 credits for 30 seconds
OmniHuman 1.5	10.8 credits per second of audio, capped at 60 seconds

On Avatar Presenter the voice-over is charged as well as the avatar render, because the tool synthesises the speech from your script — both parts are quoted together before you confirm. The exact price is shown before you generate, and failed generations are refunded. New accounts start with 10 free credits and no card, which is enough to try a face generation and see whether the persona you described is the one you want. The pricing page lists the plans and per-credit rates if you want to work out the cost of a series.

Consent and honesty

Two rules cover almost every problem here. First: only drive a photo of a real person if that person has agreed. That applies to a colleague, a client, a family member and especially to anyone whose photo you found rather than took. Second: do not present a generated presenter as something it is not. A synthetic spokesperson describing your product is fine and common. A synthetic spokesperson presented as a doctor, a customer, or a person with credentials they do not hold is a different thing, and in advertising it is the kind of claim regulators take an interest in. Label synthetic presenters where the platform or jurisdiction you are publishing on requires it, and never write a credential for an AI face that a real person would have to earn.

Frequently asked questions

Does the Spokesperson tool use my own face?

No. Spokesperson generates a face from the description you give it — gender, age, ethnicity, style and details. To build a talking video from your own photo, use Avatar Presenter and upload a clear front-facing picture.

Will the same AI presenter appear in all my videos?

Yes, if you save it. The generated face is stored in your browser and can also be saved to My Library as a reusable character, so every later video uses the identical face rather than a new similar-looking one.

Can my avatar speak in my own voice?

Not through Avatar Presenter, which uses stock voices. Voice cloning exists only inside Creator Reel, and it is limited to your own voice or one you have written permission to use.

How long can a digital twin video be?

It depends on the route. Avatar Presenter on the Kling models handles up to five minutes of audio, OmniHuman 1.5 is capped at sixty seconds, and a Spokesperson video on the video models is built from clips of four to fifteen seconds.

Is it legal to make a video with someone else’s face?

Only with their consent. The tools require a consent confirmation for uploaded photos for exactly this reason. Cloning a real person’s likeness without permission is a platform-rules problem and, in many places, a legal one.

Start with Spokesperson if you want one consistent presenter to front a whole channel, or go straight to Avatar Presenter if the video only works with your face in it. The first face generation is cheap enough to test on the free credits.