Skip to main content

Overview

The Episode Metadata Generator creates publication-ready content for your podcast website (Sanity CMS). It analyzes the episode transcript and PRF to generate:
  • Episode title (with episode number and guest name)
  • Short description (1-2 sentences for previews)
  • Long description (“On this episode…” content)
  • 3 key takeaways
  • Timestamped show notes
Metadata generation uses an agentic AI workflow with real-time progress updates. The AI analyzes your transcript to extract the most valuable content.

Generation Workflow

1

Navigate to episode

Open the episode detail page for the episode you want to generate metadata for.
2

Ensure prerequisites

  • Transcript must be uploaded
  • PRF generation recommended (provides better context)
  • Episode number and guest name must be set
3

Generate metadata

Click the “Generate Metadata” button in the episode header or metadata section.The AI will:
  1. Analyze the full transcript
  2. Extract key topics and insights
  3. Create timestamped segments
  4. Generate title variations
  5. Write descriptions optimized for SEO
4

Review generated content

A modal displays the generated metadata:
  • Title: Episode number, guest name, and topic hook
  • Short Description: Concise 1-2 sentence summary
  • Long Description: Full “On this episode…” paragraph
  • Key Takeaways: Exactly 3 bullet points
  • Show Notes: Timestamped list of topics discussed
5

Edit if needed

All fields are editable. Refine the content to match your brand voice.
6

Push to Sanity

Click “Push to Podcast CMS” to save the metadata to your Sanity episode document.This creates or updates the episode page on your podcast website.

Metadata Fields

Title

Format: Episode {number}: {Guest Name} - {Hook} Example:
Episode 348: Mark Baker - Breaking the Tech Debt Doom Loop
The title hook captures the most compelling angle from the conversation.

Short Description

1-2 sentences optimized for:
  • Social media previews
  • Email subject lines
  • Episode list pages
Example:
Mark Baker reveals why most IT organizations are trapped in a vicious 
cycle of technical debt. He shares the proven framework his team used 
to break free.

Long Description

A full paragraph (150-250 words) for the episode page. Structured as:
  1. Opening hook - Most interesting insight
  2. Context - Guest’s background and credibility
  3. Discussion topics - What you covered
  4. Call to action - Why listeners should tune in

Key Takeaways

Exactly 3 bullet points that:
  • Highlight actionable insights
  • Use specific examples or data points
  • Lead with the benefit or outcome
Example:
• The "doom loop" pattern that keeps 73% of IT teams stuck in 
  reactive mode
• Why governance models fail without executive buy-in (and how to get it)
• Mark's 3-phase transformation framework that reduced incidents by 40%

Show Notes

Timestamped list of topics discussed. Format:
[00:00] Introduction and Mark's background at Acme Corp
[05:23] The technical debt crisis in enterprise IT
[12:45] Breaking the doom loop: phase 1 - visibility
[18:30] Getting executive buy-in for transformation
[25:15] Real results: 40% reduction in incidents
[31:00] Advice for IT leaders starting the journey
Show notes are AI-generated but approximate. Verify timestamps against your actual episode audio before publishing.

Progress Tracking

During generation, watch the progress indicator:
The AI reads through the full transcript to identify:
  • Key topics and themes
  • Guest credentials and background
  • Most compelling insights
  • Natural segment breaks
Duration: 10-15 seconds

Pushing to Podcast CMS

Once metadata is generated and reviewed:
  1. Click “Push to Podcast CMS”
  2. The system:
    • Looks up your podcast episode by episode number
    • Creates a new episode document (if doesn’t exist)
    • Updates existing episode (if already exists)
    • Preserves any manually added content (audio files, images)
The push operation is non-destructive. It only updates metadata fields, leaving other episode content (like audio URLs) intact.

Best Practices

Before Generation

  • Edit the transcript - Remove filler words, fix names, clean up rough spots
  • Generate PRF first - Provides richer context for the AI
  • Set episode number - Required for proper CMS organization
  • Set guest name - Ensures accurate attribution

After Generation

  • Review the title - Does it capture the most compelling angle?
  • Check takeaways - Are they specific and actionable?
  • Verify timestamps - Approximate times should be close but may need adjustment
  • Match brand voice - Edit descriptions to sound like your show

SEO Optimization

  • Title: Include guest name and key topic (60-70 characters)
  • Short description: Front-load the most interesting insight
  • Long description: Use natural language, include relevant keywords
  • Show notes: Use descriptive phrases, not just topic labels

Regenerating Metadata

Need to start over?
  1. Click “Generate Metadata” again
  2. The AI creates fresh content from the transcript
  3. Previous metadata is replaced (local only - not pushed until you save)
If you’ve already pushed metadata to Sanity and customized it there, regenerating and pushing again will overwrite those changes.

Troubleshooting

”Transcript required” error

Metadata generation requires a transcript. Upload one first.

”Episode number required” error

Set the episode number in the episode metadata section before generating.

Timestamps don’t match audio

Timestamps are AI-generated estimates. Verify against your actual audio:
  1. Play your episode audio
  2. Note actual timestamps for key topics
  3. Edit the show notes to match

Title too long

Edit the title to 60-70 characters for optimal SEO:
  • Shorten the hook phrase
  • Use guest first name only (if clear who they are)
  • Remove episode number prefix if needed

Takeaways too generic

Edit to make them more specific:
  • Add data points or statistics
  • Include the guest’s name
  • Lead with the outcome or benefit

Build docs developers (and LLMs) love