video-to-md

Overview

Transcribes a video file, summarises the transcript with an agent, and writes the result to Markdown. The command extracts audio with ffmpeg, chunks it for the Whisper API, and generates a transcript summary when possible.

Usage

ouro video-to-md [--whisper-model <model>] [--agent-model <model>] [--ffmpeg <path>] <video> [output.md]

Key Options

• --whisper-model <model> – Override the OpenAI Whisper model (default whisper-1).
• --agent-model <model> – Override the summary agent model.
• --ffmpeg <path> – Provide an explicit ffmpeg binary.

Inputs & Outputs

• Inputs: path to the video file plus optional output path (defaults to .ouro/video/<stem>.md). Environment variables such as OURO_VIDEO_MAX_CHUNK_BYTES and OURO_VIDEO_SEGMENT_SECONDS fine-tune chunking.
• Outputs: Markdown containing run metadata, a summary section (if generated), and a fenced transcript block.

When to Use

• Capture knowledge shared in recorded demos, town halls, or video briefings.
• Provide text artifacts for downstream documentation or search.

Follow-up Checklist

• Remove temporary audio segments if they contain sensitive audio (the command cleans up temp files once complete).
• Feed the Markdown into ouro docs refine or ouro requirements refine when it contains actionable information.