Studio — Podsmith

Sources

One URL or block of text per source. The episode may only assert what these support.

URLs — one per line Pasted text or Markdown — optional

0 characters

Shape of the episode

Format

Target minutes

Lands within about ±15%.

Hosts

Disfluency —

Backchannels, interruptions, self-corrections. The strongest lever on how human it sounds.

Listener assumed to be

Groundedness threshold

Below this, no audio is rendered.

Style direction — optional Review the draft script before any audio is rendered Verify the rendered audio against the script (ASR diff, re-renders bad segments) Close with a spoken AI-disclosure line

API key

Held in this browser only, and sent as a bearer token.

Remember in this browser

Nothing running

Submit sources and the job will appear here, stage by stage.

The pipeline runs ingest → plan → script → ground → tts → post → verify → finalize. The groundedness gate sits before any audio spend, so a poorly supported script fails cheap.