How do I create a custom avatar in Synthesia?
In Synthesia, go to Avatars, click 'Create avatar,' then record via webcam or upload a 1-5 minute video of yourself following the on-screen script. You must also record a live consent video with an on-screen passcode. Processing takes about one business day. Personal avatars are available on Starter (3), Creator (5), and Enterprise (unlimited).
Why — the first-principles explanation
A custom (or 'personal') avatar works by giving the AI a reference of how your specific face moves when you talk. That's why the process centers on a short, clean video of you speaking: the model needs to see your face from a consistent angle, well-lit, with a clear mouth, so it can later re-animate that likeness to say any new script.
The quality rules all serve that goal. A single continuous take avoids stitching mismatched lighting or angles. One person in the foreground keeps the model from learning the wrong face. Hands away from the face and a simple background stop clutter from confusing the mouth and edge detection. Follow these and the avatar looks like you; ignore them and it looks off.
The consent video is the other pillar, and it's not just paperwork. Synthesia requires the same person from the footage to record a live consent statement with a randomly generated on-screen passcode, and pre-recorded consent is rejected. This exists to prevent someone from cloning a face they don't own, the core defense against non-consensual deepfakes. It's why celebrity or stolen footage gets blocked.
Once submitted, the avatar isn't instant, generation takes roughly a business day because building a re-animatable model of your face is heavy computation. After that, your face becomes a reusable avatar you can pair with any voice on your plan.
An example that makes it click
Think of it like getting fitted for a custom mannequin of yourself. The tailor films you turning slowly in good light so the mannequin matches your real proportions, that's your 1-5 minute avatar recording. If you're half in shadow or waving your arms, the mannequin comes out lumpy, which is why the lighting and 'hands down' rules matter.
Then, before the shop will make a mannequin of your face, they ask you to hold up a card with a random code and say 'yes, this is me, go ahead', that's the consent video. It stops anyone from ordering a mannequin of a stranger. A day later, your look-alike is ready to 'wear' any outfit (any voice and script) you give it.
How to do it
- Open Synthesia and click 'Avatars' in the left menu, then 'Create avatar.'
- Choose 'Create Your Personal Avatar.'
- Pick your creation method: record with your webcam or upload a video file.
- For upload, submit a single continuous 1-5 minute clip (.mp4, .mov, or .webm, up to 2GB) with one person, good lighting, hands away from face, and a simple background.
- For webcam, select your camera, microphone, and language, then record following the on-screen script.
- Record the required live consent video, reading the on-screen passcode; pre-recorded consent is not accepted.
- Submit and wait about one business day for Synthesia to generate your avatar.
Key facts
- Personal avatars require a single continuous 1-5 minute video with one person, good lighting, and a plain background.
- A live consent video with an on-screen passcode is mandatory; pre-recorded consent is rejected.
- Supported upload formats are .mp4, .mov, and .webm, up to 2GB.
- Avatar generation takes approximately one business day.
- Personal avatar limits: Starter 3, Creator 5, Enterprise unlimited (as of 2026-07).
AI avatar video creation for corporate training and comms.
Affiliate link — we may earn a commission at no cost to you.▶ The 60-second explainer (script)
How do you create a custom avatar in Synthesia? Start by clicking Avatars in the left menu, then Create avatar, and choose Create Your Personal Avatar. Now you either record with your webcam or upload a video. Either way, the AI needs a clean reference of your face speaking, so the rules matter: a single continuous take, one to five minutes, one person in frame, good lighting, hands away from your face, and a simple background. Follow those and the avatar looks like you. Next comes the consent step, and it's required. You record a live video reading a random passcode shown on screen, confirming it's really you. Pre-recorded consent won't work. This is how Synthesia blocks people from cloning faces they don't own. Then submit and wait about one business day while it builds your avatar. After that, your face is a reusable avatar you can pair with any voice. Personal avatars come with Starter, three, Creator, five, and Enterprise, unlimited.
What authoritative sources say
People also ask
Do I need consent to create a personal avatar?
Yes. You must record a live consent video reading an on-screen passcode. The same person must appear in both the footage and the consent clip, and pre-recorded consent is rejected.
How long does it take to make a custom avatar?
Recording takes only a few minutes, but Synthesia's processing to generate the avatar takes about one business day before it's ready to use.
What are the video requirements?
A single continuous 1-5 minute take, one person in the foreground, good lighting, hands away from the face, a plain background, and a visible mouth. Files can be .mp4, .mov, or .webm up to 2GB.
Which plans include custom avatars?
Starter includes 3 personal avatars, Creator includes 5, and Enterprise includes unlimited. The free plan also allows a limited number for testing.