VocalStack
Adds a vocalstack.transcript field type and a "Transcribe audio" document action. Editors click Transcribe next to an audio file; the transcript arrives with progress, survives a reload, and can be stopped. It's saved on the document as editable text plus timed sentences with speakers, ready for captions, search or a click-to-seek player. Your VocalStack API key stays on your server: a small handler (a Next.js route or any Fetch-compatible host) talks to VocalStack, and the Studio only talks to the handler. Billed per second of audio by VocalStack; the plugin is free and MIT-licensed.