A show that can be re-cut after it is recorded.

Most podcast audio is not the conversation. It is the intro, the ad read, the correction, the sponsor swap, the localised edition — all the parts that get rewritten after the episode is already recorded, and all the parts a meter charges for every single time.

107 ms

first audio, p50 on the production API

The parts that change after recording

An ad read swaps when the sponsor does. A correction lands the week after publication. A back catalogue gets a new intro when the show rebrands. Each of those is a re-render of existing episodes, priced by length on a meter and priced at nothing here.

The host voice can be cloned from the real host, so the inserted segment is in the same voice as the conversation around it rather than an obvious seam.

Two voices, one room

Film

Episodes are weekly and scripts get re-cut. Neither of those is a cost input here.

What the API does here

One request returns one complete WAV — an ad read or an intro, ready to drop into the edit.
The host voice is a request parameter, cloned from ten seconds of the real host with no enrolment step.
Emotion and prosody per request, so a correction reads as a correction and a cold open reads as a cold open.
Unmetered characters, so re-cutting a back catalogue is a decision about time rather than budget.

Notes — an engineer's checklist

01Can we clone the host so inserts match the show?

Yes — five to ten seconds of the host speaking is enough, and the reference rides in the request. Cloning requires a consenting speaker; every clip we synthesise also carries an inaudible provenance watermark applied at generation.

02Is this good enough to replace a human read?

That is your judgement to make on your own script, which is why the console on the home page runs the production API on whatever you type. We publish the measurements and the audio rather than an adjective.