Video Transcript Writer

What it does

Pulls the audio track from an embedded or self-hosted video, transcribes it, and formats the result both as a caption file (WebVTT or SRT) attached to the player and as a readable transcript block published beneath the embed. This covers viewers who can't hear the audio and gives search engines indexable text a video alone never provides.

Install

This skill isn't packaged in an installable registry. It's a plain SKILL.md file you can copy directly. Copy it from the SKILL.md tab and save it at the path for your assistant:

Use it in your AI assistant

Claude Code

.claude/skills/video-transcript-writer/ (or ~/.claude/skills/video-transcript-writer/ for all projects)

GitHub Copilot (VS Code)

.github/skills/video-transcript-writer/

OpenAI Codex

.codex/skills/video-transcript-writer/

Cursor

.cursor/skills/video-transcript-writer/ (or ~/.cursor/skills/video-transcript-writer/ globally)

Compatibility

Any WordPress site with embedded or self-hosted video. Requires the agent to access the video's audio track, or an existing auto-caption source to refine.

The complete SKILL.md for this skill:

---
name: video-transcript-writer
description: Use when a user wants a video transcribed, captioned, or made accessible, or mentions video accessibility, closed captions, or a video with no text alternative.
---

# Video Transcript Writer

## When to use
A post embeds a video (YouTube, Vimeo, or self-hosted) that has no captions and no transcript, or the existing auto-generated captions are clearly wrong.

## Procedure
1. Locate the video source: self-hosted file, oEmbed, or platform embed, and check whether a caption track or auto-transcript already exists.
2. Transcribe the audio, or clean up an existing auto-transcript, correcting misheard names, product terms, and punctuation.
3. Format the corrected transcript as a caption file (WebVTT or SRT) with reasonable line timing, and attach it to the player where the platform supports it.
4. Publish a readable transcript block below the video for users who prefer text and for search indexing.
5. Note any non-speech audio that matters to understanding (laughter, a demo sound, an alarm) as a bracketed cue in the captions.

## Guardrails
- Flag low-confidence stretches (crosstalk, poor audio) for human review rather than guessing at words.
- Don't publish a transcript that hasn't been checked against the actual audio, auto-transcripts alone are not reliable enough to ship as-is.