Repeatable business and learning videos
Synthesia fits structured training, onboarding, policy updates and explainers whose script, scene template, brand rules and review path repeat across versions.
Decision guide · AI avatars and dubbing
Synthesia is useful when a stable script and visual template must become repeatable training, onboarding or explanatory video without filming every update. This guide shows how to complete one real minute, choose Basic, Starter, Creator or Enterprise, control regeneration cost, review scenes and voices, work with guests, and separate stock-avatar licensing from custom-avatar consent before delivery.
Use Basic to build and generate one representative minute, but include download and handoff in the test because downloading is a Starter feature. Preview every scene before generation, then review the generated avatar version and correct only the affected scene. Choose Starter for downloadable, logo-free regular production; Creator for measured 30-minute capacity, API, interactive or multi-avatar needs; Enterprise only when live team editing, SSO, SCORM, brand governance or organization-scale localization is required.
Synthesia fits structured training, onboarding, policy updates and explainers whose script, scene template, brand rules and review path repeat across versions.
Pause when subject experts still dispute the content, learners need a live demonstration, paid advertising is planned with stock avatars, or no one owns final review.
Build the intended scene, preview voice and assets, generate it, collect comments, correct the version, download or share it as planned and count all regenerated seconds.
Synthesia is a scene-based AI video platform designed around scripts, avatars, voices, templates and review. A scene combines the words, speaker, language and visual layout; the service previews the draft and generates a finished avatar video. It does not decide what learners need, validate claims, clear media rights, replace detailed footage editing, approve accessibility or own the final publication process.
It turns approved text and layouts into presenter-led video using stock, synthetic or consented custom avatars and supported voices.
Teams can change a policy line, scene or language while retaining the same visual structure, review comments and published-version history.
Draft preview plays voice, assets, music and transitions but not avatar lip movement; publishing rights also differ between stock and custom avatars.
A small policy correction starts another presenter recording and edit handoff.
The affected scene is updated, regenerated and returned to the same review path.
Stakeholder notes arrive across messages and cannot be tied to a scene or version.
Guests comment on the video, replies are resolved and generated versions remain selectable.
Monthly minutes are planned without counting revisions.
Every generated second, regenerated version and delivered minute is recorded.
Choose a real section, approve its factual objective, script, examples, names, pronunciation, visual evidence, audience and required accessibility.
Split the script by idea, assign avatar, voice and language, preview speech, assets, music and transitions, then collect comments before generation.
Inspect avatar motion and lips, subtitles, brand and pacing in the generated file, revise the affected scene, then test download, LMS or sharing delivery.
| Plan | Price | Billing | Market |
|---|---|---|---|
| Basic | $0.00 | monthly | US |
| Starter | $29.00 | monthly | US |
| Starter | $18.00 | monthly equivalent, billed annually | US |
| Creator | $89.00 | monthly | US |
| Creator | $64.00 | monthly equivalent, billed annually | US |
Last checked: 2026-09-02 · This result is an estimate. Check the official checkout before buying.
Its strongest fit is a repeatable information product, not a one-off promise that an avatar replaces every kind of filming.
The editor combines a scene list, canvas and script box. Each scene can carry an avatar, voice, language, text and media, while comments and generated versions create a traceable review path. This suits training and operational communication where a controlled template matters as much as the presenter.
Before the trial, open the web app in the latest Chrome; the current supported minimums are Chrome 133 and Edge 134. Desktop requirements start at Windows 10 or Server 2016, macOS Catalina 10.15, or the listed 64-bit Linux releases. Self-serve users can use selected features such as AI Playground in a mobile browser, but the full video editor requires a desktop or laptop; Enterprise mobile use is not supported. Do not treat a successful mobile prompt as proof that the editing workflow fits.
Keep the boundary explicit: approved learning objective, script and source assets enter; a reviewed video version leaves. Subject expertise, factual sign-off, accessibility, final platform packaging, complex footage work and durable source storage remain outside the avatar generator.
Verified US prices and credits are planning data; local checkout and the active contract control other markets.
Basic is free with 1,200 credits per month and up to 10 video minutes, but downloading is listed as a Starter addition. Starter is USD 29 monthly or USD 18 monthly equivalent when billed annually, with 1,200 monthly or 14,500 annual credits, up to 10 minutes monthly or 120 annually, one editor and three guests. Creator is USD 89 monthly or USD 64 equivalent annually, with 3,600 monthly or 44,000 annual credits, up to 30 minutes monthly or 360 annually, one editor and five guests plus API, interactive and multi-avatar features. Enterprise is custom.
A self-serve generated second uses two credits. Unused credits reset rather than roll over, and self-serve users cannot top up mid-cycle. Creator is not a multi-editor plan merely because it adds guests; live team editing, SSO, SCORM, broader brand controls and organization-scale features belong to Enterprise.
A draft preview reduces avoidable errors, but avatar lips and the final generated performance still require post-generation review.
Break the final script into short scenes by idea. Preview the voice, pronunciation, visual assets, music and transitions. The scene preview does not animate avatar lips, and AI voices are nondeterministic, so record the selected language and voice and expect the generated version to need a final performance check.
Invite reviewers while the structure is still cheap to change. Resolve factual, brand and language comments before generation, render the representative minute, inspect every avatar and caption, then change only the affected scene. Test the actual download, shared page or learning-platform handoff before duplicating the project.
For delivery, generate the video before choosing the file. The current guide lists MP4 at Full HD 1080p (1920×1080), WAV audio, XLIFF for translation, and SRT or VTT captions. An interactive video with buttons or branching cannot be downloaded as MP4: preserve interactivity through a website or LMS embed, SCORM or a share link. If the buyer requires a file, duplicate the video, remove interactive elements, regenerate it and deliver the non-interactive MP4.
Ten or thirty plan minutes describe first-pass capacity; the approved minutes left after changes describe production value.
Record the subscription share, every generated second, every regenerated version, discarded scenes, reviewer time and final delivered duration. Effective Synthesia cost per finished minute equals the project share of Synthesia spend divided by delivered minutes. Keep script writing, subject review, translation, accessibility and downstream editing as separate labour costs.
Self-serve video generation consumes two credits per second, so one generated minute uses 120 credits. When an edited video is regenerated, the credit charge is recalculated from the updated full duration, not only the sentence that changed. Short scenes and pre-generation comments reduce waste, but unused plan credits do not accumulate into the next period.
Owning the final training message does not make every avatar, media asset, promotion channel or person's likeness interchangeable.
Synthesia permits stock and synthetic avatars for uses such as training, product pages and organic social publishing, but the published licence excludes paid social advertising, paid programmatic promotion and TV without permission. Custom-avatar videos follow the agreement with the represented person and require free, informed consent and applicable legal age. Decide the distribution channel before selecting the avatar type.
The customer owns Customer Data and grants Synthesia a limited licence reasonably necessary to operate, maintain and improve the service, address support or security, meet law or follow written permission. Synthesia states that customer inputs and outputs are not used to pre-train its AI components and that fine-tuning with Customer Data requires written customer instruction. The customer still remains responsible for lawful inputs and publication.
Use one approved script and delivery target; avoid a borrowed ranking of which avatar is universally more realistic.
Compare Synthesia and HeyGen with the same script, language, avatar type, preview and correction loop, reviewer roles, distribution licence, delivered-minute cost and output destination. The useful difference is which workflow constraint each service removes, not an unsupported realism score.
Use recorded screen demonstrations or a human trainer when the learner must see a changing interface, physical procedure, authentic performance or live judgment. Use slides with voice-over when a presenter adds little value. Synthesia earns a place when scenes, governance, comments and repeatable localization reduce update cost without weakening accuracy or permissions.
Yes. Basic includes 1,200 monthly credits and up to 10 video minutes. Use it to test construction and generation, but downloading is listed as a Starter feature, so include delivery in the decision.
Start with Basic. Choose Starter for downloads, logo removal, regular 10-minute production and three review guests. Choose Creator only for measured 30-minute, API, interactive, multi-avatar or additional guest needs.
Self-serve video costs two credits per generated second. Regenerating an edited version recalculates usage from the updated full duration; unused credits do not roll over and self-serve top-ups are unavailable.
Not under the published standard stock and synthetic avatar licence. It allows many training, product and organic uses but excludes paid social, paid programmatic promotion and TV without permission.
Synthesia states that Customer Data inputs and outputs are not used to pre-train its AI components. Fine-tuning with Customer Data requires written customer instruction under the governing agreement.
Comparisons