script-rip API

script-rip turns recorded speech into a usable transcript. Give it a YouTube link, an audio or video file, or a recording; it returns text with speakers separated and, where possible, named. It does not analyse, summarise, translate or edit — it hands a finished transcript to whatever does.

Two ways to consume it

Poll. Submit a run, read GET /v1/runs/{id} every couple of seconds until status is done, then fetch the transcript in the format you want. Nothing to host.

Register a destination. Tell script-rip an https:// URL once. Every finished run is POSTed there, signed, and retried for fifteen hours if your endpoint is down. Your program fetches the transcript with its own key when the webhook arrives. This is how bullshit.doctor integrates.

The two links that matter

• Quickstart — zero to a transcript in under ten minutes.
• API reference — every endpoint, every field, generated from the same declarations the API is built from. The same document is downloadable as openapi.yaml.

Conventions, once

Base URL https://<host>/v1 — on this deployment, <host> is script.rip, and every example below writes it as <host>. HTTPS only. Every request is authenticated with Authorization: Bearer $SCRIPTRIP_KEY. JSON in, JSON out, snake_case field names. Every id is a hyphenated UUID; every timestamp is RFC 3339 in UTC with a Z; durations are float seconds. Every failure is one envelope. Unknown request fields are rejected; unknown response fields must be tolerated, because adding one is how the API grows without breaking you.