November 2024 Realism is limiting

Realism is limiting and fictional is a hard aesthetic to realise.

The incident generators function. Voice experiments produce convincing material. Character structures are emerging. Yet something is unresolved. The outputs are plausible but overly stable too close to documentary reconstruction. I've also been playing with introducing noise in the system. I'm not sure why, I vassiate between thinking it's introducing a moment of distance and a sense of unreality to thinking it's like the fake lens flare on digital video, a weird skeuomorphism that draws attention to an absence.

In general the image generation models were are rapidly, producing increasingly polished and convincing visuals - well not quite really, theres still an obviousness to AI generated imagery, partly in the aesthetics of it, that smoothness and flatness, but also in the difficult to pin down pervasive unreality - either through silliness and fantasy or some kind of excess of predictability.

I've returned to some aspect of deep time, looking at ancient oceans, theres a lyrical mapping to them. notes gathered over years concerning oil histories, trade routes, early bitumen use, and prehistoric coastlines began reconnecting with the project.

Technically, experimentation continues with voice generation, layering different tonal registers: institutional, poetic, speculative, administrative. Some outputs fail completely, becoming unintentionally comic or overly theatrical. Others produce moments of unexpected coherence.

There was also uncertainty. The project risks becoming too diffuse. Without careful framing, it could expand indefinitely into history, geology, mythology, and fiction simultaneously. The challenge became learning how to let the work become stranger without losing structure.