The whole studio is one guy | Jonathan Brazeau

Summary

Jonathan Brazeau describes an AI-native creative workflow that spans RPG art, storyboards, rotoscoping, and music. He says he generated tens of thousands of images in Stable Diffusion, used Midjourney and Artbreeder for exploration, and relied on Luma AI and Suno to automate heavy production tasks like frame-by-frame cleanup and music transformation. A central theme is his “cogs and sparks” framing: humans should keep high-creativity decisions such as vision, composition, and character design, while AI should take mechanical work. He argues that with today’s tools, a small team can produce animation that previously would have required a large studio, but only if creators invest heavily in curation, consistency, and taste. Reid Hoffman widens the discussion to collective agency and the social implications of AI amplification.

Notes

• Brazeau used Stable Diffusion as a practical production tool for an RPG book, not as a novelty demo.
• He describes early image models as incomplete on key anatomy, requiring manual correction of faces and hands.
• He frames the creator-tools ecosystem as incentive-aligned experimentation, with tokens and unrestricted tool choice.
• The conversation situates Brazeau in a design-led community that originally formed around Dungeons & Dragons.
• Brazeau says his push into AI image tools began with a need for game backdrops and landscapes.
• Artbreeder was his first workaround for generating fantasy world-building assets quickly.
• He moved from Artbreeder to Midjourney and then to Stable Diffusion as the tooling evolved.
• He says his Stable Diffusion experimentation scaled to roughly 40,000 generated images.
• He emphasizes a hybrid workflow: AI generates, but the human redraws and refines.
• He compares generative AI to collaborating with another artist rather than using a passive tool.
• Brazeau adopts a “cogs and sparks” framework for dividing creative labor.
• He places UI/UX and music among the tasks where human taste still matters most.
• He says he does not want to delegate storyboarding, because vision should stay with the creator.
• He treats rotoscoping as a classic automation target once the style is established.
• He estimates that a five-minute animated film implies around 7,000 frames of manual rotoscoping.
• Luma AI is used as a labor-saving layer for repetitive frame transformation.
• He also uses Suno to rework older guitar recordings into film music.
• Brazeau thinks current tools can produce essentially any kind of animation.
• He believes small AI-native teams are approaching studio-level competitiveness.
• He describes a production workflow that starts with script, storyboard, and treatment before shooting live-action reference.
• He uses rough practical costumes as temporary input because the final look is generated later.
• He says the underlying question is what remains impossible for him to make, rather than what can be made.
• He identifies curation as the real time sink in AI art.
• He argues that taste comes from selecting one good image out of hundreds.
• Consistency requires detailed character bibles down to clothing and accessories.
• Brazeau’s fantasy western relies on palette discipline and distinctive visual signaling.
• He says AI expanded the range of ideas he can explore beyond what he can hand-illustrate.
• He frames generative models as broadening creative search space.
• He likes early Stable Diffusion for its chaotic, idea-generating behavior.
• He compares the experience to getting creative prompts from a chaotic partner.
• He likens generative remixing to Dungeons & Dragons and “chaos magic.”
• He sees image generation as probabilistic remixing rather than pure randomness.
• A single Midjourney image unexpectedly redirected the visual language of the whole project.
• He uses outlier outputs to steer the entire aesthetic direction.
• Brazeau asks Reid Hoffman about the kind of higher-order problems AI should help humans tackle.
• Hoffman responds by emphasizing collective action, not just individual tool use.
• Brazeau says he is both optimistic and pessimistic about AI, while still using it daily.
• He presents the current moment as internally conflicted rather than ideologically pure.