Script in, presenter out
You write or paste a script, pick an avatar, choose a voice and language, and Synthesia renders a video of that presenter speaking it. Templates handle slides, lower thirds, backgrounds and screen recordings around the presenter.
The consequential feature is what happens next. Change a line in the script and re-render. No studio, no presenter’s diary, no re-shoot — which is what makes this economically different from filming rather than merely cheaper.
Custom avatars are available on higher tiers: a recording session produces a presenter who is your own colleague or spokesperson rather than a stock face.
The economics that actually justify it
Worth doing the arithmetic explicitly, because it is the whole argument and it only works at certain scales.
A filmed training video costs a studio, a presenter, a crew and an edit. Updating it means repeating most of that. Localising it into eight languages means eight versions of the same cost.
With Synthesia, the marginal cost of the ninth language is a script translation and a render. The marginal cost of an update is an edit and a render. The saving is not in making the first video — it is in making the fortieth and in updating all of them.
Which is why it makes sense for an organisation with a compliance library in six languages and no sense for someone making one video a quarter. If your content does not change and does not need localising, the case largely disappears.
Writing for a synthetic presenter
The commonest reason a first attempt disappoints is not the avatar — it is the script.
Written prose and spoken prose are different. Long sentences with subordinate clauses work on a page and collapse when read aloud by anything, human or synthetic. A synthetic presenter has less ability to rescue an awkward sentence with emphasis and timing, so it exposes weak writing rather than covering it.
Practical rules: short sentences, one idea each, and read the script aloud yourself before rendering. Anything you stumble over, the avatar will stumble over more visibly.
Where the illusion breaks
Avatars are convincing in a talking-head frame and stop being convincing when asked for more. Gestures are limited and repeat. There is no walking, no interaction with objects, no genuine spontaneity. Above a minute or two, viewers notice.
This is a scope boundary rather than a defect. Synthesia is for information delivery — compliance, onboarding, product explanation, internal announcements — where a presenter’s job is to be clear rather than compelling.
For anything where performance carries the message, film a person. And where a video’s purpose is to convey warmth or sincerity — an apology, a bereavement announcement, a message about redundancies — a synthetic presenter actively undercuts it, and using one is a judgement error rather than a cost saving.
What it costs to set up
- A browser and an account. Nothing local.
- A paid plan. Free access is a demo rather than a working tier; this is enterprise-shaped software with pricing to match.
- Video minutes are metered, so long libraries need planning.
- For a custom avatar: a recording session with the person, their written consent, and a higher tier.
- A script that reads aloud well.
Disclosure, and when it stops being optional
Synthesia’s content policy is deliberately strict, and its avatars are licensed with consent from the actors who performed them — a meaningfully better position than tools that clone from found footage.
The question that falls to you is whether your audience should be told. For an internal compliance module, nobody is deceived and nobody cares. For customer-facing marketing, a synthetic presenter presented as a real employee is a small deception, and audiences respond badly when they discover it rather than being told.
Some jurisdictions are moving toward disclosure obligations for synthetic media. Deciding your own policy now is cheaper than retrofitting one.
Who buys it and why
Large organisations with training and compliance libraries in many languages, where the localisation economics are decisive.
Product teams producing feature explainers that go stale with each release, where re-rendering beats re-filming every quarter.
HR and internal communications, where content is frequent, unglamorous and needs to exist rather than to win awards.
It is a poor fit for marketing that depends on personality, for creative work, for emotionally weighted messages, and for small teams who will not use enough minutes to justify the plan.
The case for it
- Updating a video is editing text — the strongest argument, and it changes how often video gets kept current.
- Localisation at scale, from one script into a great many languages.
- Consistency: the same presenter, tone and pacing across an entire library.
- No production logistics — no studio, crew, lighting or scheduling.
- Licensed avatars with actor consent, a cleaner position than cloning from found footage.
- Enterprise controls, including brand kits and reviewer workflows.
What holds it back
- The avatars read as synthetic to most viewers, particularly over longer runs.
- Expensive, and priced for organisations rather than individuals.
- Restricted expressiveness — limited gesture range, no physical action.
- Strict content policy that refuses some legitimate topics.
- Custom avatars need consent and a session, so a spokesperson leaving is an operational problem.
- Weak economics below a certain volume.
What you retain
Rendered videos download and are yours. Custom avatars stay with Synthesia and are not exportable, so the ability to produce new video of your own spokesperson ends when the subscription does.
Keep the recording session source material and the signed consent together. And keep your scripts — they are the actual content, they represent most of the work, and they are trivially portable to any other tool or to a human presenter.
Comparable tools
- HeyGen — the direct competitor, stronger on translating existing footage.
- ElevenLabs — narration only, when no presenter is needed on screen.
- Descript — the better choice when you have real footage to edit.
- InVideo AI — template-driven video without a presenter.
Compiled from Synthesia’s documentation and public sources. We have not hands-on tested this tool. Last reviewed 16 August 2026.
