Synthesia

What Synthesia is

Synthesia is a business video platform with synthetic presenters, script editing, translation and distribution. Its current offering also includes Sessions, interactive avatar conversations for practice and surveys. A generated video and a real-time conversation are different products.

The best-known workflow transforms a written explanation into narrated scenes. The user maintains the message, visual materials and updates; the avatar supplies a presentation format. This can be useful when content changes frequently or needs several language versions. Introducing a synthetic person should serve a communication purpose: explaining, guiding or introducing an activity. The appropriate format follows from that purpose rather than from the novelty of the presenter.

Reference documentation: Video platform · Sessions.

Origins and business development

Synthesia was founded in 2017 by Victor Riparbelli, Steffen Tjerrild, Matthias Niessner and Lourdes Agapito, with headquarters in London. Its development combines research into human representation with corporate communication tools. Synthesia 3.0 brought together customisation, dubbing and interactivity; subsequent offerings extend avatar conversation.

On 26 January 2026 it announced a GV-led $200 million round at a $4 billion valuation. This describes a dated financing transaction, not educational quality or translation accuracy. Platform selection benefits more from checking how a team retains an approved script, corrects versions and withdraws obsolete material than from using valuation as a proxy for suitability.

Reference documentation: Company and founders · Series E announcement,26January2026 · Synthesia3.0.

Avatars and performance models

The editor distinguishes Express-3, Express-2, Express-1 and Legacy families. Outfit, space, pose, shot and supporting-image customisation depends on the avatar. Background and camera changes are not identical across all choices.

Select the presenter after defining tone and composition. A close shot suits a short explanation; a screen or diagram needs sufficient space of its own. Keep framing and title placement consistent across languages. If the avatar changes, review pacing, gestures and the relationship with visual elements again. Identical text does not ensure an equivalent presentation when the performance changes, particularly where a gesture directs attention towards a specific object or label.

Reference documentation: Avatar management.

From document to editable sequence

The platform can start videos from scripts, documents, presentations or links. A draft organises material into scenes for subsequent review and editing. The input supplies a basis for an explanation rather than an obligation to narrate every sentence of a document.

A useful workflow identifies audience and purpose, separates ideas by scene and decides what needs speech and what needs imagery. Mark information that must remain literal, such as a contact or internal term. Check that summarisation has not removed an exception. Adjust duration and transitions, retaining the relationship between each scene and its approved source so later updates remain traceable.

Reference documentation: Video platform.

Screen recording and motion graphics

AI Screen Recorder provides a Chrome extension that turns narration into an editable script and divides recording into scenes. Motion Graphics generates resources such as checklists, charts and timelines synchronised with the text.

Real screen capture suits explaining where an option is located; a graphic can show a sequence’s structure. Combine them according to what viewers need to learn. Prepare a demonstration account without personal data, hide notifications and inspect visible labels. When an interface changes, updated narration is insufficient: buttons, paths and screenshots also require review. Graphics should communicate verified relationships while preserving units and terminology.

Reference documentation: Screen Recorder · Motion Graphics.

Translation, dubbing and language review

Dubbing translates speech, not text embedded in the picture. It supports transcript review, translations through XLIFF and optional lip synchronisation. Adaptive duration adjusts playback to accommodate narration; the original-duration option preserves video speed.

This distinction matters for demonstrations: slowing an image can change how the timing of an action is understood. Review the complete version and consider explicit editorial pauses instead. For Spanish, Catalan and English, prepare a glossary of acronyms and names, checking narration, titles and captions separately. One voice’s quality does not establish that another will retain the same emphasis or pronunciation. Timing should remain a deliberate editorial decision.

Reference documentation: Synthesia — AI dubbing.

Downloads and delivery files

Plans with downloads provide 1080p MP4, WAV, XLIFF and SRT or VTT captions. Basic and trials do not include downloads. Interactive videos cannot be exported as an MP4 retaining their buttons and logic.

The destination determines the format: MP4 for linear playback, separate captions for a compatible player and XLIFF for translation exchange. Check that the caption version matches the delivered audiovisual file. Changing one sentence can shift later timings. Before distribution, review both playback with sound and the experience without audio, retaining the script used for final generation. Keep the files together so a later replacement does not accidentally combine mismatched versions.

Reference documentation: Synthesia — Download video.

Interactivity and assessment

Pro includes buttons and branching; quizzes require Enterprise. In a scored quiz, the results component communicates completion and pass or fail. Questions configured as practice do not send their responses through SCORM.

Design interaction around the decision you need to check. Choosing a communication channel for a concrete situation provides more information than clicking continue. Define the correct response and the explanation provided after an error. Test every path, including those returning to an earlier scene. A completed path establishes that its endpoint was reached; its educational meaning depends on the questions and criteria defined for that activity.

Reference documentation: Interactivity.

SCORM, publishing and maintenance

Enterprise exports SCORM 1.2 and 2004. The package embeds Synthesia’s hosted player rather than providing a self-contained offline copy. Changes appear after republishing; saving a draft does not update distribution. Documentation states that SCORM and ordinary embeds do not resume from where the viewer left off.

Before integrating a series, test one module in the actual campus: launch, completion, scoring where applicable and returning after exit. Retain evidence of that test. If offline access or persistent resumption is required, this dependency can determine the choice. Downloaded MP4 files require manual replacement when content changes, so distribution records should identify every destination.

Reference documentation: Publishing · SCORM export.

Sessions: conversational practice and surveys

Roleplay Sessions supports avatar conversations for practice and feedback. Survey Sessions conducts conversational interviews and produces a report of themes and sentiment. These are available functions distinct from video playback.

As a proposed OSH application, a practice scenario could rehearse communicating a concern to a team supervisor. Define a fictional situation and observable criteria: describe the issue, ask for clarification and confirm the next step. Do not interpret automatic feedback as a personality diagnosis. A conversational survey is not equivalent to a validated psychosocial assessment method; responses, data handling and interpretation need an agreed purpose and appropriate professional oversight.

Reference documentation: Sessions.

Plans and connections to other systems

The video offering distinguishes Basic, Starter, Pro and Enterprise, with different credits and functions. The API platform separates video creation, dubbing and interactive avatars. An editing subscription should not be confused with programmatic integration terms.

Choose by listing concrete deliverables and requirements: downloads, translation, quizzes, campus use, collaboration or generation from another system. Consider who will maintain templates and correct errors. Automating versions can reduce repeated work, but changes to a source document need review before replacing material already using an approved template. A repeatable production process should retain a human approval point appropriate to its content.

Reference documentation: Plans · API platform.

Avatar consent and licensing

Photo Avatars require a photograph, consent video and matching identity; the represented person must create them. Stock avatars and voices have paid-promotion restrictions. Custom-avatar uses also depend on the agreement with the individual.

Separate image permission, script approval and distribution permission. An employee may accept an internal presentation without accepting a different public campaign. Record who can use the avatar, for which messages and for how long. When the person changes role or leaves the organisation, review active pieces and the continuing permission. This record should be understandable to the colleagues who inherit responsibility for the material.

Reference documentation: Synthesia — Photo avatar moderation · Video licensing.

Technology, training and data

Synthesia describes specialised avatar-performance components trained with commissioned, licensed or public performances. It states that customer data, including inputs and outputs, is not used for pretraining; specific fine-tuning requires written customer instructions under the relevant agreement.

Uploaded documents still need control. For an internal explanation, supply only the necessary version and review appendices, names and information unrelated to the objective. Define project access and file retention. A process video can use fictional examples instead of personal records that do not help explain the task. The training commitment and the organisation’s own information-management decisions address different parts of the workflow.

Reference documentation: AI Governance Practices.

Reception, public uses and relationships with actors

Graphika linked Synthesia avatars with high confidence to Wolf News videos circulated by Spamouflage in late 2022. Its report describes deceptive third-party use. Current policy prohibits disinformation campaigns and restricts political and news uses of stock avatars.

Relationships with performers also form part of product development. In March 2025 Synthesia announced an option pool for selected actors and controls to remove their avatars from new uses. As an earlier creative example, Messi Messages combined multilingual personalisation with Lay’s and received a Bronze Lion at Cannes in 2021.

These examples concern different uses and debates: authorised personalisation, working arrangements and misleading news identities. A convincing presentation needs an audience that understands who is communicating and which elements are synthetic.

Reference documentation: Graphika:DeepfakeItTillYouMakeIt · Acceptable Use Policy · Talent experience programme · Messi Messages campaign · Cannes Lions campaign record.

Practical applications in OSH

In occupational safety and health, it can help turn an approved explanation into induction or awareness content. Its potential application is maintaining consistent narration and preparing language versions of the same message, with professional review of both the technical content and its adaptations before use. The format alone does not demonstrate training effectiveness.

It can support induction videos, explanations of internal channels and introductions to learning modules. Choose material benefiting from clear narration that can be maintained through a controlled script. Changes to a contact, address or instruction require updating every affected version.

An avatar does not replace the person answering questions or checking understanding. Within training, video can be one resource in a process including appropriate interaction, practice and assessment. Define that role before investing in a complete series.

Worked example and result checks

Example: create a welcome video explaining how to report an incident and where to find site safety information. Use a validated script, check contact details and have each translation reviewed by someone competent in the language.

1. Prepare an approved script identifying the actual channel, when to use it and where to find further information. Avoid unnecessary personal details.
2. Separate narration, on-screen text and screenshots. Give these elements a shared version reference.
3. Generate a trial and listen to names, acronyms and contacts. Correct pronunciation and pace before producing the language set.
4. Review each translation alongside titles and captions. Check that obligations, exceptions and recipients remain unchanged.
5. Test the final file without sound and in the intended player. Confirm that viewers can find the actual resource described.

This is a proposed example. Creating or viewing the video does not by itself demonstrate completion of training.

Limitations and conditions of use

An avatar can convey authority that the content does not deserve. Identify its synthetic nature, review pronunciation and captions, and do not replace a supervised practical demonstration with narration. Understanding and acquired competence still need to be checked through appropriate methods.

An avatar resembling a person requires attention to identity and permission. Synthesia sets specific consent requirements for personal avatars; someone else’s likeness should not simulate endorsement.

Fluent speech can still mispronounce an acronym or change the emphasis of a negative instruction. Review the complete video rather than only the script. Keep a human contact visible and explain the synthetic nature of the resource where needed to avoid confusion.

Reference documentation: Synthesia — Photo avatar moderation · Synthesia — Download video.

When to choose it and how to complement it

Synthesia fits scripted explanation with a virtual presenter. Consider ElevenLabs if narration alone is needed. HeyGen offers another avatar and audiovisual localisation workflow.

Test a paragraph containing proper names, numbers and an exception. Check correction time in each language and what happens when one approved sentence changes. Maintain a version inventory so obsolete files can be withdrawn from distribution channels and links can point to the correct material.

Frequently asked questions

Does an avatar represent an expert validating the content?

No. It is a presentation method. Identify responsibility and script review separately.

Does dubbing also translate on-screen titles?

Not necessarily. Visible wording requires treatment and review independent of the audio.

Does a video equal completed training?

Not by itself. The objectives, activities and checks defined for the particular training must be met.