Ben Hansford on AI Filmmaking, a Talking Raccoon and the Work Beneath a Prototype

Commercial director and USC professor Ben Hansford joins Reid Hoffman and Parth Patil to discuss AI in filmmaking, visual effects and comics. Hansford describes AI as a crew that can carry out a mission, while emphasizing the creative direction and attention required from the person using it.

The episode introduces a token grantee program offering selected creators $1,000 a week in tokens, freedom to choose tools and access to custom agents. Patil says he met Hansford through a Hollywood Slack community when few people he knew in Los Angeles were experimenting with AI.

Hansford remembers trying early generative tools before DALL-E and Midjourney became mainstream. The experiments were messy: he says he burned out a GPU and infected a computer. A mutual friend introduced him to Patil, who helped him explore newer systems.

As a commercial director, Hansford found that generated images and improved image manipulation helped him prepare visual treatments and win work. He traces his interest in storytelling to childhood, when he was a neighborhood dungeon master.

While teaching directing at USC, he began showing students new AI tools each week. He says their treatments improved and some won internal awards. He describes the development of AI teaching at the school, though his claims about being first or oldest are not independently established here.

Patil asks how the entertainment industry’s attitude has changed. Hansford says that in his experience conversations have shifted from hostility to urgent questions about applying AI. He points to pressure on budgets and distribution and to younger audiences changing their habits; this is his assessment, not a comprehensive industry survey.

Hoffman asks whether opinions remain mixed. Hansford sees a generational difference in his own students and professional contacts: some students jump into new tools while some established professionals resist them. This is an observation from his experience, not a rule about age or skill.

Hansford says an advertising project involving AI won a Cannes Lion, even as people argued with him that AI-assisted work could not win awards. He also recalls demonstrating AI visual effects to students and being challenged on why a green screen was needed at all.

A student suggested using depth and luminance mattes to isolate performers instead. Hansford took the exchange as a reminder that new practitioners may question assumptions that experienced creators take for granted. Patil sees a similar dynamic among programmers.

Asked what he can make now that was too costly or difficult before, Hansford points to commercials using AI imagery, including a spot aired during the Sugar Bowl. He says visual quality has reached a point where he is comfortable using it on television.

A second example comes from his love of comics. He is building BloqBloq as a place to create, share and explore comics using AI-assisted tools. Hansford argues that there are far fewer professional illustrators than potential storytellers; his specific comparison with the number of NBA players is his estimate and is not verified in the episode.

He says the platform lets people arrange panels and word balloons, adjust lettering and add sound effects, then share the result. Building it has also required work beyond image generation: mobile interfaces, databases, security, invitations and publishing.

Hansford calls this the “iceberg of creation.” A prototype is the visible part, but connecting a database, publishing code and securing a service are among the less visible tasks. He says he sometimes has Claude write code and Codex check it, or reverses the roles.

Hoffman asks about a “talking raccoon.” Hansford says an unannounced client project raised questions about facial capture, fur and visual-effects pipelines. Instead of discussing feasibility indefinitely, he decided to test whether existing AI tools could produce a prototype.

He acted out the raccoon’s lines and used LivePortrait and ComfyUI for facial animation, along with Move AI for body motion. He says that within about 30 minutes he had a talking raccoon that performed his lines and movements. The example is a prototype, not a comparison of final production quality or cost.

The speed of the test changed how he thinks about iteration: a director can move quickly from wondering whether an idea works to seeing a version of it. Hansford says the experiment also made him consider creating his own talking-animal show.

Patil compares new motion-capture techniques with the work of performer Andy Serkis. Hansford argues that expertise in performing for a medium becomes more valuable, not less. He recalls T-Pain discussing how a singer adapts performance for an Auto-Tune sound.

For the raccoon’s small eyes, Hansford says, he had to exaggerate eyebrow and head movements. He compares learning to perform for these systems with an actor learning the difference between a stage performance, a close-up and a wide shot.

Hansford describes an animated space project in which he performs multiple roles and transforms his voice with ElevenLabs. He imagines a small troupe developing characters and iterating rapidly before other performers take over the roles.

Asked what filmmakers should learn, he says they need familiarity with AI and the ability to work across changing formats and budgets. He argues that declining budgets predate the current AI debate; his conclusion about the causes of job losses is his perspective, not established by the interview.

Hansford resists calling AI merely a tool. A hammer cannot build a bird box overnight on its own, he says, while an agent can work on a software task and another can check it. He compares an agent to a crew, although the result still needs human review and may not satisfy the director.

His analogy is Jack Nicholson performing under Stanley Kubrick’s direction: a director can guide a performer but does not control every action. Hansford uses the comparison to argue that creative direction matters when working with systems that produce non-identical results.

Asked about opposition to AI in creative work, Hansford recalls a student who arrived intending to protest his class. He invited the student to sit through a project presentation and discuss it afterward.

One student used a recording of a deceased grandmother as narration for a personal story, recreating scenes with AI. Hansford says the projects looked and felt distinctly individual despite using the same tool. The protester praised the students’ work but still objected to what Hansford was doing.

Hansford wishes critics would look more closely at the personal work filmmakers are making. His example illustrates his view of the creative potential; it does not resolve wider concerns about labor, consent or training data.

He asks Hoffman what framework could improve a polarized public conversation about AI. Hoffman suggests that people need to experience benefits directly, mentioning possible assistants for medicine, legal documents, education and creative expression as examples of areas to explore.

Hoffman also emphasizes amplification: even where technology replaces some tasks, he wants attention on new capabilities it can give people. He sees Hansford’s class as an example of expanding creative possibilities. These are the speakers’ views of the future, not guarantees about outcomes.

Hansford says AI initially helped him win directing work, but now people can find it hard to understand someone doing several different things. He asks how to communicate that breadth without appearing unfocused.

Patil answers that a person can do more, but still cannot do everything. He describes configuring agents to pay attention to his sleep and to review their own work before interrupting him. His aim is to reserve limited attention for decisions needing a human perspective.

Agents can produce more work, Patil says, but that also creates more material to review. He prefers reviewing a finished episode or graphic novel to spending all his time tuning intermediate prompts. Protecting attention remains a challenge he says he has not yet solved.

Hoffman thanks Hansford and Patil for the conversation about creativity and directing new systems. The episode ends with their thanks and production credits.