You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Any client can now ask for a small copy of a talk break to send on by text message: GET /segments/{id}/audio?rendition=share answers AAC in an .m4a, mono at 64 kbps, so a minute is about half a megabyte where the rendered wav is three. The station makes the copy once, through the new transcode provider (the bundled audio analyzer), keeps it in its own directory (SEGMENT_SHARE_DIR, media/share-copies in the container) and sweeps it nightly once nobody has asked for it in "Keep copies listeners share for" (render.shareCopyDays, seven days by default; there is no setting that keeps them forever). Without the query the route answers exactly what it always did. A station with nothing that can make a copy answers 503 to the request and nothing else changes. Every SDK's getSegmentAudio takes the optional rendition.
A new plugin capability, transcode, makes a small copy of a piece of audio for somebody to send on rather than for the station to air. The bundled audio analyzer answers it through a new /transcode endpoint on the analysis sidecar, which encodes with ffmpeg's own AAC encoder (mono at 64 kbps by default, so a minute of speech is about half a megabyte). It has its own provider choice, "Make copies to share with" (render.transcodePluginId), so a station can join audio with one plugin and make copies with another; leave it empty and the first plugin that can is used, as for every other provider. Nothing asks for a copy yet. The console and the desktop app name the new capability on plugin cards and in Providers.