How to Make an AI Avatar Video With Synthesia
Synthesia turns a script into a video with a lifelike AI presenter, and its cheapest paid plan starts around $18/month billed yearly for 10 minutes of video. You pick an avatar, paste the script into a slide-based editor, and render an MP4 in about the time it takes to read it.
What is Synthesia?
Synthesia is a UK-founded video generation platform, started in 2017 by Victor Riparbelli, Steffen Tjerrild, Lourdes Agapito and Matthias Niessner. It is the most established name in avatar-based corporate video, used heavily by enterprise L&D teams for onboarding and compliance training.
The core product is not a video editor. It is a script-to-video engine: text on the left, rendered presenter on the right. That constraint is deliberate. Companies do not need creative control over their tenth “how to file an expense report” video; they need it produced in an hour by someone in HR.
Synthesia reported over 50,000 business customers by 2025, with names like Heineken, BSH and the UK’s NHS in its public case studies.
How do you create a video with Synthesia?
The workflow fits into five steps and the editor actively steers you through them.
- Choose “Create video” and either start blank or pick a template. Templates carry the scene count and slide style, which is what makes corporate videos look consistent.
- Select an avatar from the library or upload a custom one on higher tiers. Over 230 stock avatars are available across 140-plus languages.
- Type or paste your script. Each scene becomes one paragraph; the avatar speaks exactly the text in that scene.
- Add slides, images, screen recordings, or a brand kit with your logo and fonts.
- Generate. Rendering takes roughly real-time to double real-time; a five-minute video lands in 8-15 minutes.
Editing after generation is cheap. Change one scene’s script and re-render only that scene, not the whole video.
The habit that saves the most time: write the script as five to seven short scenes rather than three long ones. Long scenes make slides static and give you nowhere to cut when a stakeholder wants a change.
What does Synthesia cost in 2026?
| Plan | Price | Video minutes |
|---|---|---|
| Free | $0 | 3 minutes, 36 minutes of video lifetime (trial) |
| Starter | About $18/month yearly | 120 minutes/year, 1 editor seat |
| Creator | About $64/month yearly | 360 minutes/year, 1 personal avatar |
| Enterprise | Custom | Unlimited-ish, SCORM export, SSO, custom avatars |
The minute math is the thing to check before you buy. A 5-minute training module per month is 60 minutes a year — Starter covers it. Ten modules a month is 600 minutes and pushes you to Enterprise quickly.
Custom avatars require filmed consent and a recorded session; that is an Enterprise-tier feature and takes about a week to process.
How good is the avatar output?
On a well-lit, front-facing talking-head clip, it is convincing enough for internal training and increasingly for customer-facing product explainers. Lip sync for English and the main European languages is tight. The weak spots are predictable: rapid-fire technical jargon, brand names the model has not seen, and any script with heavy emphasis or emotional variation.
The tell is not the face — it is the delivery. AI presenters do not pause for effect the way humans do. Writing shorter sentences with deliberate full stops improves perceived naturalness more than any setting in the app.
One more thing: background music is not included. Budget a library (Epidemic Sound runs about $10/month) or the videos will feel unfinished no matter how good the avatar is.
Where does Synthesia fall short?
- Predictable avatar movement. Hands and posture repeat across scenes; viewers notice by minute four in a long video.
- Minute-based billing is unforgiving. A 6-minute script eats 6 minutes of quota for one render, and re-renders of changed scenes still consume minutes.
- Limited creative range. It will not do a dynamic explainer with motion graphics; that is an After Effects or Vyond job.
- No offline or desktop version, and enterprise deployments usually want SSO plus data-residency options that only exist at the top tier.
- Language quality varies. English, German and French narration hold up; Japanese and Arabic need a native review pass.
- Stock avatar library is large but visibly AI. For public marketing, a custom avatar or a filmed piece still reads better.
Who should use it?
L&D teams producing onboarding, compliance and product training at volume. Marketing teams localizing one explainer into eight languages without booking eight voice actors. Anyone who needs a video out the door by Friday and has the script already written.
Do not use it if you are making brand films, if your content depends on real human emotion, or if you publish fewer than four videos a year — a single contractor shoot costs less than a year of subscription.
Quick Summary
- Synthesia is a UK avatar video platform (founded 2017) with 230+ stock avatars in 140+ languages.
- Paid plans start near $18/month yearly for 120 minutes of video; custom avatars are Enterprise-only.
- Render time runs roughly one to two times the video length; scene-level re-renders save quota.
- Script in short scenes and short sentences — that is what makes AI delivery sound human.
- Best for corporate L&D and localization; wrong tool for brand films and motion-graphic explainers.
