The core message of the session was that AI video tools can feel “magical” because they radically shorten the path from imagination to execution, but the human creator still supplies the “soul”: taste, story, intent, character, and emotional meaning. After an opening reference to seeing the first Dutch AI-generated or AI-made movie, “Operation Amsterdam,” the speaker connected filmmaking and AI agents, arguing that there is something worth exploring between cinematic characters and agent design. Rinte then framed his presentation around his personal experience of working with AI in filmmaking “day and night” and repeatedly being amazed by new capabilities, while emphasizing that AI “can never become human” because it lacks a beating heart and lived experience.

He illustrated that renewed sense of creative magic with several concrete production examples. A childhood memory of using a Minolta VHS camera set up the idea that AI has restored the feeling of being an eight-year-old discovering audiovisual storytelling for the first time. He then showed how toy cars could be used as references for action-scene previsualization, how a screenshot from Google Maps could become a simulated Amsterdam drone shot, and how a six-year-old nephew’s drawing could be animated into a living character. These examples supported his claim that barriers are disappearing: instead of needing crews, permissions, actors, contracts, and post-production infrastructure, creators increasingly need “a great idea and a computer.” He described this as the democratization of filmmaking.

The strongest proof point was his satirical Paris tour-guide film, inspired by news images of Paris on fire and a real viral clip of a guide continuing a tour amid chaos. He cast an AI-generated guide named Elisa, prompted the scenes, and said the finished piece went worldwide viral: about 10 million views within 24 hours and more than 100 million views by the end of the week, all made from his office between Saturday afternoon and Monday. From there, he showed the rapid progress of AI video using the “Will Smith eating spaghetti” benchmark, moving from poor 2023 results to realistic footage and one-prompt fantasy scenes such as a spaghetti monster fight.

The latter part of the provided transcript focused on control and workflow. Rinte showed character sheets based on LinkedIn images, location references from the venue, hybrid filming methods where a performer controls an AI version of himself, and camera-control workflows using Blender. He demonstrated asking an LLM such as ChatGPT, Claude, or Gemini to operate Blender, create a 3D reference from a toy soldier photo, animate camera movement, and export an MP4 that could guide AI video generation. He closed this segment by pointing to Google DeepMind Genie 4 as a glimpse of real-time interactive worlds, predicting that brands such as Carlsberg or pharmaceutical companies could eventually host online worlds where customers explore, ask questions, and interact with products or information.