Synthesia

There is a version of this company's founding story that starts with a research paper. Victor Riparbelli had moved from Copenhagen to London in 2016 to explore what deep tech actually looked like up close. He was consulting for the UK government on VR and AR, thinking broadly about where the next wave was coming from, when he encountered a paper by Professor Matthias Niessner demonstrating that AI could synthesize realistic lip movements synchronized to arbitrary audio. His reaction, in his own words: "I just felt like I saw magic. I saw the technology, and I was like: this is going to change everything we know about media production."

He found Niessner, found Niessner's collaborator Professor Lourdes Agapito, brought in Steffen Tjerrild to run operations, and with Lasse Korsgaard as an early technical contributor, co-founded Synthesia in 2017. Ninety-eight European VCs said no. Mark Cuban said yes. The rest, as they say, is a $4 billion company.

The Problem Synthesia Solves — And Why It Scales

Traditional video production has a fundamental unit economics problem: it costs roughly the same to produce one video as it does to produce ten, because the fixed costs — studio rental, camera operators, talent scheduling, editing — barely change with volume. For a company that needs to produce training videos in 14 languages for 65,000 employees across 150 countries, this is not a production problem. It's a math problem. Synthesia solves the math problem.

Type a script. Pick an avatar. Click generate. Get a broadcast-quality video in minutes. No studio. No camera. No coordinating with a talent agency in eight countries. And if the compliance team changes one sentence in the script six months later, re-render just those three seconds — at no marginal cost.

What's New in Synthesia 3.0

Synthesia has been iterating quietly and shipping loudly. The current platform — internally called Synthesia 3.0 — represents a substantial leap from its origins as a talking-head video tool:

• AI Playground — integrated access to Google Veo 3.1 and Sora 2 for generating cinematic b-roll clips directly inside the Synthesia editor; each 8-second clip costs 48 credits and is available on all plans including Free
• Express-2 Avatars — Synthesia's highest-fidelity avatar generation tier; full-body, expressive, with natural gestures and micro-expressions that eliminated the uncanny valley complaints of earlier versions
• Video Agents — coming soon; AI-powered interactive video that can have real-time conversations, answer viewer questions, and adapt content dynamically based on viewer responses
• Courses — a full eLearning module builder built natively into Synthesia, enabling L&D teams to go from script to complete interactive course without an LMS or third-party authoring tool
• AI Copilot — a generative AI script assistant that writes, edits, and localizes video scripts from a brief or a URL, eliminating the blank-page problem for non-native English copywriters
• AI Dubbing — upload any video and receive a lip-synced, voice-matched dubbed version in 160+ languages; available on Starter and above
• Interactivity — add clickable CTAs, branching narrative paths, and quiz modules to any Synthesia video without leaving the editor

The Numbers Are the Argument

"Komatsu replaced voiceover training slides with Synthesia avatar videos in 14 languages — and 80–90% of viewers now watch the full video." — Komatsu L&D case study
"Coursera used Synthesia to translate instructor-led courses with preserved teacher voice and emotion, lifting course completion by 25%."
"Würth Group cut video translation costs by 80% across 80+ countries."

These are not anomalies. They are the expected outcome when you remove the cost and logistical friction from multilingual video production. Synthesia's enterprise customer base — which includes Zoom, SAP, Bosch, Merck, Heineken, Amazon, Google, BBC, and Reuters — is growing because the ROI is visible in the first project.

Security & Ethics at Scale

Synthesia takes a harder line on content safety than virtually any other AI video platform. Every avatar creation requires verbal consent verification from the person being digitized. An automated moderation layer plus a human review layer screen every video before publishing. The company was one of 200+ signatories to the AI Safety Summit commitments in the UK, and Victor Riparbelli regularly advocates publicly for AI governance frameworks that distinguish between legitimate synthetic media and malicious deepfakes. SOC 2 Type II, ISO 42001, and GDPR compliance are standard across all paid plans.