The Working Assembly's AI art for Untold

The Working Assembly, Untold, 2026-09-02: The agency has built a bespoke AI illustration system that generates painterly images which respond to what users say in the voice-journalling app Untold.

This isn’t a case of feeding prompts into a public image generator to cut costs. Founder Jolene Delisle says the team began with a human-made visual reference and then layered a custom AI agent onto a strict set of creative rules and guardrails. "For us, it wasn’t really AI versus an illustrator or about replacing an illustrator," she says.

Practically, that means the visuals are designed to evolve as people use Untold: not a fixed asset library but a dynamic visual language that adapts to expressed emotions, tones and themes. The Working Assembly tuned texture, lighting, composition and other details so outputs match a specific, crafted look rather than random variations.

Why this matters

For music and audio fans, the project points to how creative tech can make listening and speaking into richer, personalised experiences. If apps can translate voice into evolving, bespoke artwork, similar techniques could be used for album art that reacts to your playlists, live show visuals that shift with the crowd’s energy, or podcast players that generate imagery tied to an episode’s mood.

Delisle acknowledges the ethical and aesthetic tension around machine-made interpretation of something intimate like a voice. But she frames the AI as a tool: designers craft the inputs, set the boundaries and curate results, rather than being replaced by them. The role shifts from making every asset to directing and refining an ongoing system.

Where designers fit in

The Working Assembly’s approach underlines a likely future: human authorship will look less like producing every single frame and more like defining a living visual grammar. For creators and fans who care about authorship, that means studios and designers who can write the rules and edit outputs will remain essential.

Whether you’re into album art, podcast aesthetics or interactive music experiences, Untold’s work shows how generative systems can plug into audio-first products to deliver visuals that feel personal and immediate.