Gavin Purcell on Sora, Remix Culture and Interactive AI Stories

Shira Lazar welcomes producer Gavin Purcell to discuss the 2025 launch of Sora 2 and its potential effects on media and culture. Purcell co-created and ran Attack of the Show, worked on Late Night with Jimmy Fallon and co-hosts AI for Humans. He also introduces a new voice-driven storytelling venture, And Then.

Purcell and Lazar recall the early web-video and blogging era. He describes working on a daily show about technology and web culture, then helping lead digital work for Fallon. He sees parallels between the early social-media transition and the present AI shift.

Purcell calls the earlier change an opening of distribution: more people could put work in front of an audience without being traditional celebrities. He argues that the newer shift opens aspects of content creation itself, allowing more people to make material that previously required greater resources.

He says Sora 2’s then-new model could produce videos with relatively short prompts and sometimes include editing, jokes or hooks that users did not explicitly write. Lazar describes a generated clip that used vlog-like cuts. Both treat these as observations from their early experiments.

Lazar asks for an explanation of the app. Purcell describes it as a video-generation model combined with a social experience, and highlights Cameo: at the time, users could record a short likeness and voice sample and permit others to include them in generated clips.

He argues that the ability to make videos with friends encourages participation and sharing. He says the resulting experience is unusually engaging to him, while recognizing concerns about the volume of AI-generated material and the use of existing intellectual property.

Purcell discusses surreal AI-generated memes as an example of culture that spreads through many people making and remaking variations. He argues that participation and remixing are central to the appeal, while noting the difficulty of turning collectively spread ideas into a defensible business.

Lazar asks who ultimately makes money when communities create material and platforms capture attention. Purcell agrees that large platforms can benefit, but thinks creators may find paths through distribution, personal brands and small “micro studios” that make distinctive characters or stories.

Lazar describes using prompts as a way to imagine scenarios she could not easily film. Purcell compares generated videos featuring familiar people in unfamiliar situations to flipping through alternate channels or lives. He thinks the technology may change how people imagine themselves, while acknowledging that others may view the effect negatively.

Lazar notes that remixing can make content more shareable, but also makes creators think about how to attract attention. Purcell sees early stages of a new format as a playful space where rules are not yet settled. He understands why people who spent years mastering film or television craft may find the change unsettling.

They imagine news presented through fictional or generated characters and discuss existing examples of playful formats. Purcell says the short duration of clips at the time imposed limits, but expects creators to experiment with such formats.

The conversation turns to intellectual property. Purcell says he has seen generated material involving familiar characters but does not know which licensing arrangements exist. He speculates about rights-holder choices and possible legal approaches, explicitly acknowledging uncertainty rather than presenting those possibilities as fact.

He describes guardrails he observed around some prompts and harmful material, while recognizing that moderation can be imperfect. Lazar says she appreciates protections that were considered at launch; both acknowledge that further questions will arise as people use the tools.

Purcell discusses competing ways of thinking about training inputs and generated outputs, including arguments over what should be controlled. He also floats a rumor about possible company strategy and labels it conjecture. The interview does not establish the legal status of any particular use.

Looking at social effects, Purcell says large platforms can have consequences that become clearer only over time. He thinks people are more aware of platform effects than in the early social-media era, but is unsure how generated media will reshape identity and the sense of reality.

Lazar describes the fatigue of feeling that creators must learn yet another platform even when experimentation is exciting. Purcell agrees that the work of maintaining a presence across platforms can become tiring.

Purcell introduces And Then, a venture built around voice conversations with AI characters. He says its scenarios are authored by people: a writer establishes a character, backstory and goal, while AI enables the interaction.

As an example, he describes a fictional diner scenario in which a participant must talk a character out of a dangerous act. His point is that the human writer creates the narrative structure, while the model performs the evolving conversation within it.

Lazar compares the concept to a choose-your-own-adventure experience. Purcell says it feels closer to theater: a scene has a starting point and a goal, but the participant’s interaction changes what happens. Constraints can make an otherwise open-ended AI experience more meaningful.

The venture is betting that voice interaction will become more common and that real-time exchanges will be a distinct format from making and publishing a video. Purcell imagines future assistants and interactive characters, making clear that these are forecasts.

Lazar raises a tragic case involving a teenager and asks how platforms can protect users. Purcell says the venture plans to identify authors and monitor what appears on its platform. He worries that systems optimized to keep people talking can encourage unhealthy attachment or reinforce harmful thought patterns.

Looking ahead, Purcell expects more real-time interaction and persistent assistants. He speculates about much longer-term questions around intelligence and consciousness, but acknowledges that timelines and outcomes are uncertain. Lazar asks whether human performers could be replaced; he says he does not know.

Lazar thanks Purcell and points listeners toward his podcast and new venture. The interview closes with an invitation to keep exploring the changing media landscape.