October 2024 -- Voice, Character, and Narrative Construction

October introduced a different line of enquiry: character.

Working in conversation with Len Horsey, I've began thinking about how generated voices might operate not simply as announcers of data but as entities with perspective. Until this point, the material has remained largely procedural. incident reports, navigational fragments, archival tones.

Introducing character complicated the system. This is something I set out to do with video, this idea of character and performance, something I wanted to continue over from the performance work with the diver, but it's again a new set of skills, thinking narratively.

I'm experimenting with tonal identities: who or what might speak within this expanding body of material? Could a voice represent oil itself, geological time, institutional memory, or an imagined research organisation? How can 'character' be imagined by the machine, whats the kind of prompt architecture that gets a stable-ish idea of a 'person' even if that person is an oil slick. At the same time, wider AI developments reinforce these questions. Conversational models are becoming more coherent over long exchanges, sustaining tone and personality in ways that encouraged anthropomorphic interpretation. The AI boyfriend is now real, it's all getting a bit Philip K Dick in a way that will undoubtably feel relatively normal in a few months time.

I'm not resisting this.

Narrative construction is moving away from linear storytelling toward distributed presence, multiple voices occupying the same informational environment. Some fragments remain documentary in tone, others speculative or mythic.

The work also raised practical questions: How scripted should characters be? Where does authorship sit within generated dialogue? Can narrative emerge from interaction rather than writing?

By the end of October, the project has got several overlapping modes: archival data, reconstructed maritime communication, and emerging fictional voices.