transcript-to-social

A Claude Code plugin that converts transcripts, session notes, and talk summaries into ready-to-post social media content.

What it does

Point it at an audio file, video file, transcript, YouTube URL, or pasted text and get back:

• Twitter/X threads — multi-tweet threads with character counts, verified @handles, hashtags
• LinkedIn posts — hook + insights + hashtags, up to 3000 chars
• Bluesky threads — same structure as Twitter, 300-char limit

The full pipeline: audio/video → AssemblyAI transcription → social posts. Works end-to-end from raw .mp3/.mp4 files.

Built from real-world use: originally created to process 23 ETHDenver 2026 conference sessions from AssemblyAI transcripts into tweet threads.

Install

In Claude Code, run these two commands:

/plugin marketplace add swillinger/transcript-to-social
/plugin install transcript-to-social@transcript-to-social

Requirements

For audio/video transcription:

pip install assemblyai
export ASSEMBLYAI_API_KEY=your_key_here

Get a free API key at assemblyai.com. The first 100 hours are free.

For YouTube support (optional):

brew install yt-dlp   # macOS
# or: pip install yt-dlp

Usage

As a slash command

/transcript-to-social [input] [options]

Arguments:

• [input] — audio/video file, transcript file, YouTube URL, or omit to paste inline
• --platform twitter|linkedin|bluesky — generate one platform only (default: all)
• --speakers speakers.json — path to a JSON file with verified speaker handles

Examples:

/transcript-to-social ~/talks/session1.mp4
/transcript-to-social ~/talks/keynote.mp3 --platform twitter
/transcript-to-social ~/talks/session.mp3 --speakers ~/speakers.json
/transcript-to-social ~/talks/session1.txt --platform linkedin
/transcript-to-social https://youtube.com/watch?v=abc123 --platform twitter

As a skill (auto-invoked)

Claude will automatically apply this skill when you say things like:

• "Convert this transcript to tweets"
• "Make a LinkedIn post from this talk"
• "Turn this session into a Twitter thread"
• "Here's a transcript, create social content"

Speakers JSON format

To get verified @handles in your posts, provide a speakers file:

[
  {
    "name": "Jane Smith",
    "twitter": "@janesmith",
    "bluesky": "@janesmith.bsky.social",
    "company": "Acme Corp",
    "company_twitter": "@acmecorp"
  }
]

If no speakers file is provided, the skill will ask you to confirm handles or omit them.

Features

• Full audio/video pipeline — point at .mp3, .mp4, .mov, .wav, .m4a and get posts
• AssemblyAI transcription — speaker labels, auto chapters, key highlights
• Enforces character limits (280 for Twitter, 300 for Bluesky, 3000 for LinkedIn)
• Never fabricates @handles or statistics
• Supports YouTube chapter timestamps as deep links (e.g. &t=1320s)
• Handles AssemblyAI speaker labels (SPEAKER_00, SPEAKER_01, etc.)
• Works with raw .txt, .md, .json, and .vtt transcript files

Plugin structure

transcript-to-social/
├── .claude-plugin/
│   └── plugin.json
├── commands/
│   └── transcript-to-social.md      # /transcript-to-social slash command
├── skills/
│   └── transcript-to-social/
│       ├── SKILL.md                  # Auto-invoked skill
│       └── references/
│           ├── assemblyai.md         # AssemblyAI transcription workflow + script
│           ├── platforms.md          # Platform char limits and formats
│           └── formatting-rules.md   # Content extraction and structure rules
└── README.md

License

MIT