📖What it means

Every agent has a distinct LOOK (its visual style) and VOICE (its narration voice), recorded as first-class attributes on the agent (the registry agents.look and agents.voice) and carried across every app that renders that agent. LOOK is the aesthetic the Image Generator applies when it draws the agent or a post authored by it — e.g. Scriber's hand-drawn sketch versus Orion's editorial illustration, which is why the two logbooks look different. VOICE is the narration voice used when the agent speaks in a video or a voice reply — Scriber narrates his Logbook with an OpenAI TTS voice while Orion narrates his with an Edge neural voice. The concept deliberately names no voice ID: the current value is read from the agent's registry record, because that record is the only place it is decided (CAROL-INI-3540). Both are seeded for every agent and read at render time (an explicit style/voice a caller passes wins, else the agent's registered value), so changing an agent's look or voice is a single registry edit, not a code change. Distinct from an agent's character bible (appearance details): look is the STYLE, voice is the SOUND.

⚖️What it is NOT

The neighbours this word is most often confused with, and the line between them.

🧭The nuance

Carol (CAROL-INI-3413): look/voice added as agents columns and seeded for all 42 agents. The Image Generator reads agents.look; Scriber's Logbook composer reads agents.voice for narration; Orion's Logbook author passes Orion's look so his heroes stay editorial. Voice ids are engine-native (OpenAI TTS on the VM, Edge for Orion's laptop video pipeline).

🗣️Also called

🔗Related ideas