What would happen if a recorded podcast could suddenly talk back to you? ACTx486 is exploring exactly that idea, using an episode of The Joe Rogan Experience featuring Elon Musk as the stage for an unusual experiment in interactive video.
The result looks familiar at first: Musk and Joe Rogan appear to be having their original conversation. Then the viewer steps in. A question is asked, the recording changes, and a synthetic version of the person on screen responds as though the conversation has continued. The technology is impressive, but the details behind those responses are just as important as the demonstration itself.
ACTx486 turns a recorded podcast into interactive AI video
Created by Jakub Zegzulka and Karina Nguyen, ACTx486 is presented as a “research demo of a new interactive medium.” Nguyen highlighted the project on X on September 23, 2026, describing an experiment built around a simple but potentially huge shift in how people experience recorded media.
The creators explain the concept with the question:
“What if you could talk to a video?”
They say ACTx486 explores how existing media could “listen, respond, and adapt.”
Instead of opening a separate chatbot while watching a video, the concept places the interaction directly into the footage. A viewer can interrupt what is playing, ask something, and then see generated material appear within the video before the original recording continues.
The demonstration uses a real episode of The Joe Rogan Experience featuring Musk. Musk first appeared on Rogan’s podcast in 2018, on episode #1169, and has subsequently returned to the show multiple times.
That familiar source material is important to the experiment because the audience already recognizes the people, voices and environment. ACTx486 then inserts generated segments around portions of the genuine recording, creating the impression that the viewer has temporarily stepped inside the conversation.
More AI-Powered Ventures: A digital community’s ambitious plan to build a real city in Uruguay could welcome its first residents as soon as 2027. For more, read our story on Praxis’ $1 Billion AI City in Uruguay.
How the ACTx486 demonstration works
| Feature | ACTx486 demonstration |
| Main source material | A real episode of The Joe Rogan Experience featuring Elon Musk |
| Creators | Jakub Zegzulka and Karina Nguyen |
| Project description | Research demo of a new interactive medium |
| Demonstration highlighted | September 23, 2026 |
| Viewer interaction | Interrupt a recording and ask questions |
| Generated elements | Faces, voices and words in synthetic segments |
| Languages | Includes a demonstration in Italian |
| Current response speed | Generated in advance; complete interactions currently take minutes |
| Future target | Creators would ideally like responses in under six seconds |
| Affiliation | Not affiliated with The Joe Rogan Experience or its hosts |
The technology is therefore less like a normal chatbot sitting beside a video and more like an attempt to make the video itself become the interface.
AI-generated Musk can answer questions he never answered
This is where the demonstration becomes particularly striking.
In one example, the viewer interrupts the podcast and asks Musk a question. ACTx486 generates a response designed to match the visual presentation, voice and mannerisms of the person in the original recording.
That response was not actually spoken by Musk.
The project also demonstrates capabilities beyond simply making a synthetic host speak. ACTx486 can generate visual explanations, allowing diagrams and other visuals to appear while the simulated host explains an idea.
When a question is unclear, the simulated host can also ask the viewer to clarify what they mean.
The system can change the setting and continue the conversation around newly generated material. Once the interaction is over, the original recording can resume, making the viewer’s request feel like a temporary detour rather than a replacement for the original podcast.
The creators describe the viewer as becoming a “microdirector”, since the audience can influence what happens while the original work remains at the centre of the experience.
More AI Breakthroughs: A century-old WWI cipher finally gets cracked, with archival naval records confirming the decoded message. For more, check out our coverage of GPT-6 Astra’s WWI Cipher Breakthrough.
ACTx486 can switch the simulated host into Italian
Another demonstration highlights just how far the concept goes beyond conventional question-and-answer AI.
ACTx486 shows the simulated host responding in Italian. The system is designed to let viewers speak in another language and receive a response in that language while maintaining the simulated host’s voice and personality.
According to the creators, hosts can switch languages independently while the surrounding episode continues in its original language.
The generated response is designed to appear directly inside the video. That means the system adapts elements such as the host’s gaze, timing, speech and transitions so the interaction can feel native to the footage.
This is a crucial part of ACTx486’s broader idea. Rather than having an AI assistant operate alongside a video, the technology attempts to make the video itself responsive.
The Musk and Rogan responses are completely synthetic
The realism of the demonstration makes its disclosure particularly important.
ACTx486 explicitly states:
“Every clip in this post is a research demonstration built on a real episode of The Joe Rogan Experience with Elon Musk. In the generated segments, their faces, voices and words are synthetic. Nothing they say there was said by them, and nothing in these clips is a statement by either of them.”
The project also states that ACTx486 is not affiliated with The Joe Rogan Experience or its hosts.
The creators say generated portions are labelled on screen. The disclosure changes at the point where genuine footage stops and generated material begins, making the distinction visible to viewers.
That separation matters because ACTx486 is deliberately designed to make generated dialogue blend with real footage. Without a clear distinction, viewers could potentially mistake synthetic words for something that was actually said during the original recording.
The creators specifically identify this as a major issue surrounding the technology.
Why ACTx486 raises deepfake concerns
ACTx486 does not present its technology as being without risks.
Its website directly addresses the possibility that capabilities designed to make interactive media more useful could also make synthetic misinformation more convincing.
The project states:
“Dual use. The capabilities that make interactive media compelling, like reproducing a person’s face, voice, mannerisms, knowledge and reactions, can make information far more useful and falsehoods far more believable. The same capabilities also enable deepfakes.”
The Musk and Rogan example makes that issue especially clear because genuine and generated footage are intentionally combined.
ACTx486 explains:
“Real words are mixed with invented ones. A single interaction can blend the host’s recorded words, invented words in his voice, and invented words quoting things he really said, so we tied disclosure to the system’s own edit list, and the on screen label switches exactly where real footage stops and resumes.”
The creators are therefore treating disclosure as part of the underlying experience rather than an optional extra.
The technology is designed to create a seamless interaction, but the project recognizes that the more seamless the generated material becomes, the more important it is to show audiences exactly where the authentic recording ends and the synthetic material begins.
The current ACTx486 demo is not happening live
There is another major distinction between what the clips look like and what the technology currently does.
The demonstrations were generated in advance.
ACTx486 can route a request, research information, write a response and generate the resulting scene, but a complete interaction currently takes minutes.
The creators say they would ideally like responses to arrive in under six seconds. The current technology has not reached that target.
So, despite the appearance of an instantly responsive podcast, ACTx486 is not currently a system where someone can pause a live or recorded episode, ask a question and immediately receive a generated answer.
The existing clips are demonstrations of the experience the creators are trying to build.
That distinction places ACTx486 firmly in the category of a research experiment rather than a finished consumer product.
The AI can personalize elements of the video
The concept also extends beyond asking a synthetic Musk or another simulated host a question.
According to its creators, ACTx486 can maintain a profile of the viewer and use that information to personalize responses. Information that a viewer wants to keep can also be sent to their phone as a card.
The demonstrations include locations, visualized ideas and different environments. Viewers can ask for information to be represented visually, introduce objects into a scene and continue interacting with them.
Importantly, changes can persist between interactions instead of resetting each time.
An object introduced into a scene, for example, can remain there while the viewer asks additional questions. The system can also allow both people in a scene to respond to a request or let the viewer address whichever host is visible.
All of those features support the same larger concept: the video becomes the interface.
ACTx486 includes safeguards for its synthetic hosts
The developers say they have already introduced some restrictions, although they acknowledge that the safeguards remain incomplete.
They state:
“We have built some guardrails, but we need more. We prevent the host from drawing on later parts of the episode and require the system to complete its research before using it in a response.”
The system is also designed to decline personal and political questions on behalf of the real person whose likeness is being simulated.
The creators say they are continuing to explore a wider trust-and-safety framework for media that speaks as another person.
Their broader question is also explicitly about control:
“We’re still exploring how much control belongs to the viewer and how much should remain with the creator.”
That question sits at the heart of the project. ACTx486 is attempting to give audiences more control over recorded media without necessarily replacing the original work.
Musk and Rogan did not participate in the demo
The project also openly explains that it used real people’s likenesses and voices without asking them to participate in the research demonstration.
ACTx486 states:
“Likeness and source material. We used real people’s faces and voices and short excerpts from one episode, solely for this research demonstration.”
The creators say they chose the Musk and Rogan episode because both are highly recognizable. That makes the concept immediately understandable: viewers know what the original people look and sound like, making the contrast between genuine footage and generated interaction particularly meaningful.
There is currently no indication in the supplied project materials that Musk or Rogan participated in creating the demonstration. ACTx486 explicitly says it is not affiliated with the show or its hosts.
What ACTx486 could mean for the future of recorded media
For now, ACTx486 remains a research demonstration, with its website collecting email addresses for a waitlist as the creators continue experimenting with interaction, response speed and safety.
The podcast example is intended to show a broader possibility rather than limit the technology to podcasts.
The same basic approach could theoretically be used with educational videos that answer questions while viewers watch, interviews that provide additional context on demand, documentaries that branch into related subjects, or entertainment that lets viewers influence individual moments.
But the project also recognizes that greater control is not automatically the right answer for every form of media.
The creators write:
“Generative media offers infinite possibilities, but puts the burden on the user.”
They then raise the larger issue:
“But the choices made by someone else are part of what makes a work worth experiencing. We’re still exploring how much control belongs to the viewer and how much should remain with the creator.”
That tension could ultimately be just as important as the technology itself.
More Copyright Battles: A takedown dispute over video game music licensing has escalated all the way to federal court. For more, read our report on the GameChops v. Materia Music Lawsuit
Conclusion
ACTx486 offers a glimpse at a version of video where watching no longer has to be completely passive. Its Musk and Rogan demonstration shows how viewers could interrupt recorded footage, ask questions, receive synthetic responses, request visual explanations, change languages and influence elements of a scene.
But the most important distinction remains clear: the generated Musk and Rogan segments are synthetic and do not represent statements actually made by either person. The current interactions were also generated in advance and can take minutes rather than happening instantly.
For now, ACTx486 is a research experiment exploring what happens when the boundary between watching and interacting with recorded media becomes much thinner. The technology may be intriguing, but its creators are also confronting the difficult question of how much control viewers should have over works created by someone else—and how clearly synthetic content must be separated from reality.
Disclaimer
This article has been prepared based on thorough research of the sources provided in the verified material above, including the ACTx486 project website and research disclosures, information concerning The Joe Rogan Experience and Elon Musk episodes, Karina Nguyen’s September 23, 2026 X announcement, and RuntimeWire’s report on ACTx486. The article reflects the information available from those provided sources and does not imply independent verification beyond the material supplied.
Sources
- ACTx486 official project website
- Joe Rogan Experience: Elon Musk episodes
- JRE Clips: Elon Musk’s first appearance, Episode #1169
- Karina Nguyen’s ACTx486 announcement on X
- RuntimeWire’s report on ACTx486
Featured Image Credit: Debbie Rowe / Creative Commons Attribution-Share Alike 3.0
Leave a Reply