On a few selected posts, I have now added audio transcription, partly as an experiment and partly because it may prove useful for people with accessibility needs, or for anyone who would rather listen than read.

This has involved working with the ElevenLabs API, exploring text-to-speech generation, voice sample technologies, and the practicalities of turning written material into a more natural spoken format. The aim is not simply to bolt on a robotic reading of the text, but to create something that feels more considered: clear, listenable, and sympathetic to the atmosphere of the original post.

The technology around sampled and synthetic voices is becoming increasingly impressive, especially when used carefully. A voice can now carry tone, pacing and emphasis in ways that make long-form writing feel more approachable. For a site that often deals in memory, atmosphere and half-remembered cultural traces, that opens up some interesting possibilities.

At this stage, it is selective rather than universal. Some pieces suit audio better than others. But where it works, it adds another layer: not just text on a screen, but something closer to being read aloud from the other side of the room — hopefully useful, and perhaps faintly uncanny in the right places.