Victor Riparbelli and Synthesia’s enterprise route to market

A synthetic-video company can find an enterprise market by attaching the technology to a repeated, budgeted communication job rather than selling novelty. Training, onboarding and internal explanation involve scripts that change, audiences in several languages and content that must be updated. If software can make those workflows faster and more consistent while meeting security, consent and brand controls, a buyer has a reason to procure it.

That is the commercial logic behind Victor Riparbelli and Synthesia. The company did not need every corporate video to become synthetic. It needed a class of content for which studio production was slow or disproportionate and a buyer who could measure the improvement. The enterprise route also raised the standard of proof. A convincing demo could open a meeting, but deployment required administration, governance, data protection and an answer to the risks of realistic generated media.

The use case turned a research capability into a workflow

Synthesia's current company page identifies Riparbelli as co-founder and chief executive and dates the business to 2017. It describes roots across several universities and a mission focused on video communication. That record supports his current role, while also making clear that Synthesia was not a solo-founder project.

The underlying capability, generating a presenter-led video from text and other inputs, can be demonstrated in minutes. A commercial product needs a more specific answer to "what happens next?" Enterprise training and communication provide one. A subject expert can revise a script, an approved presenter can deliver it consistently, and teams can create language versions without repeating an entire physical shoot.

The choice reduces several forms of friction at once. It lowers the threshold for producing an instructional video, shortens the update cycle and creates reusable templates. More importantly, it maps to work that organisations already fund. Learning, people, compliance and communications teams have owners, deadlines and source material. The technology enters through an existing job rather than asking a buyer to invent a department for synthetic media.

This does not mean generated video is always cheaper or better. A high-stakes leadership message, dramatic campaign or complex demonstration may need performance, location, credibility or craft that the format does not provide. Product-market fit comes from identifying the repeatable subset, not claiming to replace every camera.

Enterprise value comes from revision and localisation

Traditional production cost is incurred not only on the first version but whenever a policy, product or regulation changes. A training team may need to revise one paragraph across several modules and languages. A software workflow can make that change less disruptive if the script, visual assets, pronunciation and approvals are managed in one place.

Synthesia's enterprise product material focuses on training, localisation, brand controls and collaboration. These are clues to the chosen buying case. The value is not simply that an AI avatar speaks. It is that a distributed organisation can produce, review and update a family of related materials.

Localisation deserves careful treatment. A generated language track can accelerate a first version, but a fluent reviewer still needs to check terminology, tone, pronunciation and local legal meaning. A literal translation of safety or compliance language can be wrong even when the voice sounds natural. Enterprise buyers should define which content requires in-country review and keep evidence of approval.

The economic case should be measured against the real baseline. Count scripting, studio, editing, translation, review, revision and distribution on both sides. Include subscription, training, governance and human quality-control time for the synthetic workflow. Avoid using one unusually expensive physical shoot as the comparison for every video.

The route to market follows an accountable buyer

Enterprise communication has a recognisable budget owner, but several other functions can veto a purchase. Information security asks where content and personal data are processed. Legal teams examine rights in scripts, images, voices and avatars. Brand teams need templates and approval. Employees may ask whether a synthetic presenter is identifiable and whether management is replacing a real speaker in a sensitive context.

This creates a multi-party sale. A successful pilot should therefore test more than generation quality. It should establish user roles, content approval, retention, export, deletion, accessibility and support. Procurement can stall when a demo sponsor has not involved security or legal colleagues until the end.

Synthesia publishes security practices describing hosting, access and audit arrangements and provides a Trust Centre for assurance material. Those disclosures help a buyer begin due diligence. They are not a substitute for assessing the buyer's own use case, contract, required controls and current audit evidence.

For founders, the lesson is to design the procurement path alongside the creative interface. Single sign-on, permission levels, audit logs and content governance may appear less distinctive than generation quality, yet they determine whether a large organisation can move from an enthusiastic user to an approved deployment.

Synthetic media makes consent a product requirement

Realistic generated presenters create risks that conventional slide software does not. A person's likeness or voice can be used outside the context they expected. A viewer can mistake generated delivery for a fresh statement by a real executive. A malicious user can attempt impersonation or deceptive content.

An enterprise product needs controls at several layers. The provider must establish how stock and custom presenters are created and authorised. The customer needs approval rules governing who can create a presenter, which scripts can be published and how generated material is labelled. The organisation should retain a record connecting the final video to its approved script and version.

Disclosure should reflect context. An internal software tutorial may need a simple statement that the presenter is generated. A message that appears to come from a named leader requires stronger consent and audience clarity. Content about health, employment, finance or public events deserves additional review because a polished delivery can amplify an error.

Safety controls can introduce friction, but that friction can support the enterprise proposition. A buyer is more likely to standardise on a platform if it can show that authorised use is governed and misuse is investigated. The relevant question is not whether the system makes abuse impossible. It is whether prevention, detection, response and remedy are credible for the risk.

London supported a company aimed beyond London

Synthesia describes a founding network connected to UCL and other international universities. In 2025, the company announced a larger London headquarters at Regent's Place. A London base can provide research, product and enterprise-sales talent while keeping the company near large corporate buyers and creative industries.

The market was never confined to the city. A video platform becomes more useful as it supports distributed teams and languages, so its sales, support and governance need an international scope. London's contribution is an organisational base and network, not the complete customer geography.

This distinction matters when assessing the city's AI ecosystem. Office expansion shows a commitment to a location, but it does not allocate all company output to that city or prove local economic impact. A serious assessment would examine where research, product leadership, customer work and decision-making actually occur.

What AI founders should copy, and what they should not

The transferable move is to find a frequent workflow with expensive revision, not to copy the avatar interface. Interview the people who produce and approve the current output. Map every handoff and quantify delay. Build the smallest governed workflow that improves a defined measure, then let the customer compare it with the existing process.

Choose a use case with a tolerable error boundary. An internal draft can be reviewed before release; a live public impersonation cannot be recalled in the same way. Start where human approval can catch problems and where the synthetic nature of the content does not undermine its purpose.

Do not assume enterprise logos prove broad adoption or business value. A customer may have run a small pilot, received free access or used the product in one team. Evidence should state the number and type of active users, the workflow replaced, review cost, frequency of updates and whether use continued after the initial contract.

Riparbelli's route is valuable because it links an arresting technical capability to mundane organisational work. The durability of that route will depend on whether the platform remains trustworthy and economical after the novelty fades.

Limitations and questions for publication

This profile relies on Synthesia's company, product, security and office materials. It does not include interviews with Riparbelli, co-founders, customers, employees, people represented by avatars, security auditors or procurement teams. Company descriptions establish intended features and controls, not independent proof that every deployment is secure, accessible, cost-saving or effective.

Before publication, Shoreditch Talk should review current audit evidence under appropriate access, test the content-reporting process and interview buyers whose contracts and usage can be defined. Any production-time or cost comparison should disclose its baseline and include review work. The company should be asked how it authorises custom presenters, labels generated content and responds to misuse. Riparbelli's role and the London operating footprint should be reconfirmed.