Synthesia Puts Interactive Avatar of NVIDIA's Toy Jensen at GTC Berlin
Synthesia has built an interactive avatar of Toy Jensen (TJ), the cartoon version of NVIDIA founder and CEO Jensen Huang, for NVIDIA GTC Berlin. From October 21 to 22, TJ will talk with attendees at STATION-Berlin and recommend conference sessions based on their interests.
- Synthesia built an interactive avatar of Toy Jensen (TJ), the cartoon version of NVIDIA CEO Jensen Huang.
- TJ is at STATION-Berlin during NVIDIA GTC Berlin from October 21 to 22, recommending sessions from an agenda of over 100.
- The avatar uses Synthesia's video and voice models, trained and running on NVIDIA GPUs, plus a large language model tied to the agenda.
- Off-topic questions are ended politely or redirected to the agenda.
- Synthesia says the same technology can power interactive training videos and workplace conversation practice.
A conference guide that talks back
According to Synthesia, GTC Berlin is NVIDIA's European flagship conference. It spans three days of talks, panels and hands-on workshops in one venue, and the agenda includes over 100 sessions. Instead of scrolling through the programme, attendees can walk up to TJ at Session Room 6 or between Session Rooms 3 and 4. They tell him what they work on or what they want to learn, and leave with a shortlist built for them.
Synthesia gives an example of how this works in practice. A founder looking to create the next breakaway AI app gets a different set of recommendations from an engineer building robots. Both can ask follow-up questions and receive personalized, context-aware answers, the company says.
Built on Synthesia models and NVIDIA hardware
Synthesia states that it built the avatar with its own video and voice models, trained and running on NVIDIA GPUs. Everything TJ says and does comes from the company's proprietary models and its Interactive Avatar API, packaged in a widget that can be embedded in apps or websites.
- The video model generates TJ's face, expressions and lip movement.
- The voice model generates his speech.
- A large language model matches visitors' questions to the information in the event agenda.
The scope is deliberately narrow. If a question falls outside the agenda, the avatar will either politely end the conversation or redirect it back to the agenda, according to Synthesia.
From watching video to talking with it
Synthesia describes TJ as a playful example of where business video is heading. The same technology, it says, lets employees put questions to a training video or rehearse a workplace conversation with an Interactive Avatar. In the company's words, video evolves "from something you watch to something you can talk to".
The TJ project follows other recent Synthesia announcements centred on interactive avatars. The company's blog lists the launch of Sessions on October 1, a platform where Interactive Avatars coach, interview and guide users. It also notes that Synthesia was named to Fast Company's Next Big Things in Tech 2026 list in the International category for avatars that hold two-way conversations.
For event organisers and companies, the GTC Berlin deployment shows one concrete use of the embeddable widget: a conversational guide limited to a defined body of content. For attendees, it offers a way to filter a packed agenda by personal interest rather than browsing it manually.
