Ben Relles belongs to a generation of people who learned to make video at a time when YouTube was still finding its own language. In 2007 he co-founded the channel Barely Political. The team became known for the viral parody “Crush on Obama”, which was made during the US presidential campaign and became one of the first political viral hits on YouTube. He later spent roughly a decade inside YouTube. In 2019, according to Deadline, he moved to a new innovation division focused on interactive formats and new types of content. In June 2022 he left the company and joined the team around Reid Hoffman, according to Tubefilter.
On 21 May 2026, in an exclusive report for The Hollywood Reporter, he announced a new company, Make Believe. It is an AI lab focused on interactive video formats, in which the viewer asks questions and the image is meant to respond in real time. The investor is Reid Hoffman, co-founder of LinkedIn and now an active AI investor, who himself runs the publicly available digital twin Reid AI. The first commercial partnership is with A+E Networks (the American TV media group jointly owned by Disney and Hearst) and its channel, The History Channel. The aim is to have interactive avatars of historical figures that viewers can question. The Hollywood Reporter does not disclose the size of the seed round.
What Make Believe is building
Make Believe is building technology for video in which the viewer communicates with an avatar or with the video itself, and the content adapts to their action. Relles put it this way in an interview with The Hollywood Reporter: “For 100 plus years video has been something you watched. We think part of the next era of video is something that is genuinely interactive.”
The Hollywood Reporter published specific product scenarios. In one video, a head chef answers viewers’ questions while they prepare a recipe at home. In another, a fitness creator comments on the technique of the person exercising during a workout. A guitar teacher guides a student through a specific song. The logic is the same across cases. The image is meant to respond to the viewer’s action in real time.
Relles’s positioning against Hollywood is explicit. He told The Hollywood Reporter: “A lot of the conversation in Hollywood is around how AI will make things cheaper and faster for movies and TV, and I would say for us we're more focused on what formats were impossible before AI existed.” He mentioned educational applications separately: “People are visual learners, people learn through interactivity. I see a lot of applications through science and math.”
Competition in real-time conversational video
Make Believe is entering a market that in 2026 is quickly dividing by latency and use case. In real-time conversational video, the players include Tavus, with its Phoenix-4 model launched in February 2026, HeyGen, which has moved from Interactive Avatar to LiveAvatar, and D-ID Agents. Alongside them is enterprise text-to-video and avatar video workflow, where Synthesia is strong, with 240+ avatars and 160+ languages, as well as Hour One. According to the Hollywood Reporter’s description, Make Believe falls into the first group, real-time interactive video. It is video that is meant to respond to the viewer’s question or action.
According to Tavus’s own product page, Tavus Phoenix-4 is a full-duplex model that listens and responds at the same time, with sub-one-second latency. HeyGen moved Interactive Avatar towards LiveAvatar, meaning real-time interactive avatars for face-to-face conversations. Synthesia, by contrast, is primarily an enterprise AI video platform for corporate communication, training and marketing, with an emphasis on avatars, localisation, compliance and LMS workflows, according to its product page.
Alec Lindsay, who joined Make Believe from the HeyGen team according to the Hollywood Reporter, comes directly from this category. Make Believe has not yet disclosed its technology stack, a date for commercial launch or a pricing model.
Reid AI as proof of concept
Reid Hoffman is not only an investor in Make Believe. The company’s technology builds on the Reid AI project. Reid AI is a digital twin of Hoffman which, according to a HeyGen case study, is built on a large corpus of Hoffman’s material, including twenty years of his podcasts and interviews and five books he has authored. It can answer in real time and search for answers in specific sources, depending on the question asked. In 2025 it won a Webby Award in the Best Use of AI in Video and Film category, according to the same case study and the Silicon Valley Video Summit 2025 page.
Hoffman’s relationship with Relles dates back to June 2022. Relles described his position on Hoffman’s team at the time as follows: “Reid has a general belief that every tech company needs a content strategy, and that’s a big focus of my role.” Tubefilter quoted him. Reid AI functions as a publicly traceable proof of concept for the type of technology on which Make Believe is building.
From Hoffman’s perspective, this is not a pure seed bet without history. Reid AI has already operated for three years as a public pilot of the type of technology that Make Believe is now commercialising, and it has a publicly traceable output in the form of a Webby Award. That gives the company a reference project in commercial negotiations.
The EU regulatory layer
Article 50 of the EU AI Act introduces transparency obligations for AI-generated content and takes effect on 2 August 2026. Deployers of AI systems that generate images, audio or video constituting a deepfake must inform viewers that the content has been artificially generated or manipulated. Article 50 contains an exception for “evidently artistic, creative, satirical, fictional or analogous work or programme”, where the obligation is limited to appropriate disclosure that does not hamper the display or enjoyment of the work.
For similar AI-augmented video production distributed in the EU, labelling the origin of content becomes a key question from 2 August 2026. For breaches of the transparency obligations, a penalty framework of up to €15 million or 3% of global annual turnover is cited under Article 99 of the AI Act.
For Make Believe, this means that commercial deployment of the product in the EU from summer 2026 will probably require clearly addressed AI disclosure, especially for avatars, digital twins and historical or fictional figures that may appear authentic. The European debate on Article 50 is moving in 2026 towards specific rules on labelling and detectability. The Draft guidance / Code of Practice process works with formats such as watermarking, metadata, visual labels or audio disclaimers. The final interpretation will be decisive for specific implementation.
What we know so far
Make Believe is entering the interactive AI video market with a team that has deep experience of YouTube formats. The team is made up of Ben Relles as founding CEO, Margaret Burris, who previously worked at YouTube, and Alec Lindsay from the previous HeyGen team. The first publicly announced commercial partnership is with A+E Networks and The History Channel. The technology has already been tested inside the Reid AI project.
Open questions remain the amount of funding, the date of commercial launch, the technology stack and the structure of the pricing model. It is not clear whether Make Believe will go down the route of a standalone app for creators, enterprise deals similar to the A+E partnership, or licences for existing video platforms. The Hollywood Reporter did not disclose any of these points in its exclusive report.
What matters is that Make Believe is built on the thesis that video is beginning to function as an interface. That is a shift in a different direction from making video production cheaper, which is where most AI tools are currently aimed.
Comments