Recon — Video Transcriber

A Recon contributor: during pre-processing it transcribes referenced videos — local files or YouTube links — into text. It prefers existing subtitles and falls back to Auris STT, optionally using Iris for visual context, then contributes the transcript snippet so Synth can reason about the video’s content.

Recon contributors expose no model-facing actions.

Configuration

Key	Purpose
RECON_VIDEO_TRANSCRIBER_RECON_ENABLED	Enable this Recon contributor.
RECON_VIDEO_MAX_SECONDS	Max video length to transcribe.