← Back to overview

For a voice platform · production

How do you direct a voice the way a director guides an actor?

Directing a performance, not writing a prompt — for a voice platform.

Today tone is coaxed out of a voice by rewriting the prompt or hand-tagging text with markup — closer to coding than directing.

Why now generative voice is now good enough to ship — so teams need a performance they can direct, not a prompt they rewrite.

Use cases Audiobook & podcast production, dubbing, character voice, ads.

Interactive demo ↗

The question

A prompt describes a voice once. But a performance is a hundred small choices — a whisper here, a pause there, one word leaned on. Those don’t fit into a single sentence.

The approach

Treat the script as the surface. Drop directions straight onto it — delivery on a line, emphasis on a word, a timed pause after it — the way notes land in a margin.

The feature

A palette of directions — whisper, warm, sarcastic, pause, interrupt, overlap — placed by hand. Two readings, A and B, play back to back so you can hear the difference.

Function

The direction palette

direction desk — the palette
Loading…

Delivery, focus, timing and interplay — each chip placed onto a whole line or a single word.

Function

A line, directed

Late again. The city doesn’t wait for anyone.

∿ whisper · on the line ▾ emphasis · on “anyone” △ pause · after

Directions attach to the text itself — a delivery on the line, an accent on one word, a beat after it.

Newsletter

Notes from building at the human–AI interface.

Every two weeks: new builds, notes, explorations — the interaction patterns that are working for agentic AI, and short field notes you can put to use.

Subscribe