Kwebbelkop Interviewed by His Own AI Twin — Built in One Week
Watch the live Kwebbelkop AI-twin demonstration at LEAP and DeepFest 2024.
The brief
Create a convincing digital version of a globally recognised creator, make it conversational and deploy it as a live interviewer on a major conference stage — with only days to prepare.
Building the twin
The team combined four layers into a single experience:
- Visual identity: a high-resolution likeness animated through Yepic’s real-time facial engine.
- Voice: a low-latency cloned voice designed to preserve Jordi’s recognisable tone and cadence.
- Movement: characteristic gestures, facial expressions and speaking rhythms studied from existing appearances.
- Persona: an LLM context built around Kwebbelkop’s personality, while assigning the AI the role of interviewer.
The live moment
On stage, the AI twin questioned its human counterpart in a natural back-and-forth exchange. The demonstration made a difficult technical idea instantly legible: an AI agent could have a recognisable face, voice, personality and purpose — and interact with a live audience without relying on a text box.
Why it matters
For enterprise teams, the project demonstrates rapid customisation under event pressure and the ability to join avatar rendering, voice, language-model logic and streaming into one public-facing system. The same architecture can support digital experts, brand ambassadors, hosts, trainers and customer-facing agents.
Open the project video on YouTube
The AI twin on stage
Frames from the original project footage show the live exchange between Jordi and the autonomous version built to interview him.
The AI twin introduced as a live on-stage participant.
Jordi in conversation with the real-time persona.
The system operating as an autonomous interviewer rather than a passive likeness.
The human creator and his Yepic-built AI counterpart sharing the stage.
