Victor Riparbelli on Synthesia, AI Video and the Future of Content

Synthesia co-founder and CEO Victor Riparbelli joins Harry Stebbings to discuss building an AI video company, enterprise adoption, foundation models and the future of content. Synthesia sells AI video tools, so his product, labor and market predictions carry a direct commercial interest. The interview does not independently verify its growth, customer or competitive claims.

Riparbelli announces a $100 million Series D led by NEA with existing investors participating. He recalls that in 2017 the founders' vision of generating video was difficult for investors to understand, and says roughly 80 to 90 declined before a $1 million seed round at a $5 million post-money valuation with Mark Cuban. These figures are his recollections.

The company initially focused on AI dubbing. A later attempt to raise $8 million took nine months and ended with a smaller $3.1 million round, he says; the Series A followed after the company found a sustainable product. He believes scarce capital forced the founders to focus on customers rather than pursue too many projects, although a different funding outcome cannot be tested retrospectively.

As an angel investor, he worries that too much early capital can encourage hiring and building several products before a team understands customers. Stebbings challenges the notion that product-market fit is a one-time achievement. Riparbelli agrees that growing companies need successive fits as products and markets change, while founders should own the search for the next one.

He says Synthesia has spent conservatively and raised funds to preserve flexibility and invest when opportunities are clear. He imagines a $50 billion to $100 billion company in the category. That is an ambition, not a valuation forecast supported by financial projections in the interview.

His long-term thesis is that synthetic video and audio could make visual communication as scalable as text, expanding beyond today's video-production market. He speculates that future generations may no longer read and write as the default way to communicate. This is a content-company founder's vision, not evidence that text will disappear or that video is better for every task.

Stebbings asks whether raising capital to deter competitors is a bad strategy. Riparbelli says capital can send a signal, but the priority should be building the product; spending without a purpose can backfire. He says competitors have helped educate the market and claims Synthesia benefited from having a superior product. The comparison is his own, not an independent test.

On the AI hype cycle, he says many enterprise buyers are told to pursue an AI strategy before they understand their actual needs. Pilots and new contracts can therefore look promising without producing lasting value. He considers renewal a stronger signal than the first signed contract, though renewals still require context and do not alone measure customer outcomes.

Riparbelli believes AI investment contains a bubble in which some products will fail or become features of larger platforms. He is skeptical of the phrase 'AI employees' and says impressive demos do not establish that software can autonomously handle complex business operations. He expects skilled engineers to become more productive, but the timing and extent are predictions.

He thinks text-generation models are becoming more interchangeable for many uses and that distribution and product design will matter. Asked to invest at hypothetical valuations, he picks xAI because of perceived upside, execution and the distribution of X. This is a speculative personal investment view, not a verified valuation or a recommendation.

He expects further progress from compute, algorithms and data, but rejects the idea that the most compute alone guarantees victory. For generated video, he says a central challenge is control: keeping a character consistent and specifying actions without sacrificing visual quality. These are his technical judgments, and the interview provides no controlled model comparison.

On moderation, Riparbelli says he favors approaches using community input but acknowledges difficult gray areas, including distinguishing enthusiastic cryptocurrency promotion from fraud. He says Synthesia takes a strict approach because enterprise customers do not want their avatars associated with questionable material. That is a company policy with business incentives, not proof that its decisions are always correct.

Asked whether foundation-model providers could compete directly with Synthesia, he argues customers buy an end-to-end workflow: drafting, avatars and voice, editing and distribution, not a model alone. He says the company has million-dollar-plus contracts, without detailing terms or customer counts. His claims about differentiation and quality are those of an interested vendor.

He expects both large general models and specialized models embedded in many products, while urging attention to the products themselves. He predicts a future model release will improve capabilities but may not recreate the cultural shock of the first ChatGPT launch. These are expectations at the time of the interview, not confirmed outcomes.

Riparbelli predicts digital generation will lower technical barriers to video and audio production, letting more creators realize ideas. He imagines the technical cost of making a Hollywood-style film eventually approaching zero. Stebbings challenges the near-term claim, citing AI clipping tools that still fall short of his team's judgment; neither timetable nor total production cost is established.

Riparbelli agrees that lower technical barriers do not eliminate the need for skill. He expects storytelling, selection and creativity to matter more, and thinks audiences may also value demonstrably real performances amid abundant synthetic media. These are predictions, and their effects on creators and audiences remain uncertain.

He expects more AI-generated than human-made content in five years, but says discovery matters more than raw volume. He likes TikTok's interest-based recommendations and would prefer users to have more control over their feeds. His characterization of platform behavior and his own feed is personal experience, not an independent audit.

Stebbings defends social media as a matter of individual content choices; Riparbelli agrees there can be useful material while acknowledging wasted time and negative consequences. He is optimistic about a media ecosystem with wider participation and community context. Their discussion does not establish that algorithms or community moderation reliably prevent harm.

On identity and fraud, he proposes verifying identities in some spaces and building provenance for content: when it appeared, who made it and whether it has been edited or translated. He imagines tracing a clip back to its original source. This is a design proposal; verification, privacy, access and reliability are unresolved in the interview.

He says provenance could identify an old image reused as if it showed a current event, and argues that authenticity of intent can matter more than whether production involved AI. He considers a shared record and content matching as possible components. The interview does not demonstrate a system that can reliably prove identity, consent or context across platforms.

Asked about jobs, he expects some technical production tasks to shrink or change while storytelling and creative direction gain importance. He suggests visual-effects artists could use AI to work faster and predicts more content creators overall. His illustrative productivity estimates and transition optimism are forecasts, not evidence that every affected worker will find a replacement role.

He says Synthesia began in London partly because he could not obtain a US sponsorship, but thinks the US West Coast still offers advantages. He describes European talent as potentially more loyal and questions whether average engineers in the Valley work harder. Those comparisons are personal impressions, not controlled labor-market findings.

He sees strong UK technology companies but urges policymakers to maintain an attractive environment for founders, including tax incentives and potentially support for compute. Stebbings expresses a more negative view of the recent environment. Their policy and regulatory assessments are opinions from industry participants, not an analysis of fiscal tradeoffs.

He sees London as culturally attractive to international talent, but says the local stock market needs improvement while acknowledging he is not a finance specialist. He does not consider the location of a potential listing his top concern compared with incentives for entrepreneurs. These are judgments, not a decision or announcement of an IPO.

In quick fire, he argues strategy games helped him practice rapid decisions and feedback, while recognizing they are not the same as running a real business. He describes himself as a curious generalist and says broad interests can help him connect ideas. His views on games and childhood are personal, not evidence of a guaranteed developmental outcome.

He says he has relearned not to follow competitors out of guilt or fear of missing out, and instead to listen to paying customers. Asked which investor he would choose again, he names Mark Cuban for believing in the company early. That praise comes from a founder with an investor relationship.

He says taking some secondary liquidity can reduce financial stress and help founders make decisions without all their wealth tied to one stock. He believes he might have considered selling earlier without it, a counterfactual based on his experience. He does not disclose transaction amounts or establish that secondary sales are appropriate for every founder.