Skip to main content
The PRF (Podcast Repurposing Framework) is the strategic foundation of your episode content. It analyzes the conversation, extracts key insights, and structures content for repurposing across social media and marketing channels.

​
What is a PRF?

A PRF is an AI-generated document that:
  • Identifies key themes - Main topics and discussion points
  • Extracts insights - Actionable takeaways and quotable moments
  • Structures content - Organizes information for easy repurposing
  • Highlights virality - Flags content with social media potential
  • Provides context - Summarizes guest expertise and episode narrative
The PRF serves as the “source of truth” for all downstream content generation. High-quality PRFs lead to better hooks, posts, and visuals.

​
How PRF generation works

​
Automatic generation

PRF generation starts automatically when you upload a transcript:
1

Transcript uploaded

The system receives your transcript via file upload or paste.
2

AI agent activated

The PRF agent analyzes the transcript using the configured AI model (default: Claude 3.5 Sonnet).
3

RAG search performed

The agent searches Pinecone for relevant brand guidelines, past PRF examples, and content templates.
4

PRF document created

A structured analysis document is generated and saved to the episode.
5

Status updated

Episode status changes to “In Progress” and guest status updates to “In Production”.
File reference: src/pages/EpisodeDetailPage.tsx:218-255

​
Agentic mode with RAG

YBH Pulse Content uses agentic AI for PRF generation:
Agentic AI means the system:
  • Uses semantic search to find relevant context
  • Applies brand voice guidelines automatically
  • Tracks status updates in real-time
  • Verifies facts against the transcript
During generation, you’ll see status updates:
  1. Analyzing transcript - Reading conversation structure
  2. Retrieving context - Searching brand guidelines and examples
  3. Generating PRF - Creating structured analysis
  4. Verifying facts - Cross-checking claims against transcript
File reference: src/pages/EpisodeDetailPage.tsx:223-238

​
PRF structure

A typical PRF document includes:

​
Executive summary

A 2-3 paragraph overview of the episode:
  • Guest background and expertise
  • Main topics covered
  • Key insights or controversial takes

​
Core themes

Major discussion themes with supporting quotes: 
## Theme: IT Governance as Enablement

**Key Insight**: Governance doesn't mean bureaucracy—it's about enabling teams to move faster with confidence.

**Supporting Quote**: "The best governance frameworks are invisible. Teams don't feel constrained; they feel supported."

**Repurposing Angle**: Contrast "governance as red tape" vs. "governance as acceleration" for LinkedIn infographic.

​
Viral moments

Sections flagged for high social media potential:
  • Contrarian statements
  • Surprising statistics
  • Memorable analogies
  • Quotable one-liners

​
Content opportunities

Suggested content formats:
  • Infographic ideas (data points, frameworks, comparisons)
  • Quote cards (pull quotes with context)
  • Video clips (timestamped segments)
  • LinkedIn posts (angles and hooks)

​
Editing the PRF

PRFs are fully editable using the TipTap rich text editor:

​
Editor features

  • Rich formatting - Bold, italic, headings, lists
  • Structure - Headings, subheadings, bullet points
  • Highlights - Mark key sections for easy reference
  • Comments - Add inline notes (team collaboration)
  • Version history - Sanity tracks all changes
File reference: src/pages/EpisodeDetailPage.tsx:470-474

​
When to edit the PRF

Edit the PRF to:
  • Add missing insights the AI overlooked
  • Remove off-topic sections
  • Strengthen viral moments with better framing
  • Ensure brand voice consistency
Hooks are generated from the PRF, so improvements here cascade downstream.
You can still edit the PRF, but changes won’t affect existing hooks unless you regenerate them.
If you make major edits to the PRF (e.g., changing themes or insights), regenerate Hooks and Posts to ensure consistency.

​
Regenerating the PRF

To regenerate the PRF with updated settings or a different model:
1

Navigate to PRF tab

Open the episode detail page and click the PRF tab.
2

Click 'Generate PRF' button

The button appears at the top of the tab (even if a PRF already exists).
3

Wait for generation

The existing PRF is replaced. This takes 30-60 seconds depending on transcript length.
Regenerating the PRF does not automatically regenerate hooks or posts. You’ll need to regenerate those separately if you want them to reflect the new PRF.
File reference: src/pages/EpisodeDetailPage.tsx:218-255

​
PRF approval workflow

Approve the PRF when it’s ready for downstream content generation:

​
Approval process

  1. Review content - Read through the PRF for accuracy and completeness
  2. Make edits - Fix any issues or add missing insights
  3. Click ‘Approve’ - Green checkmark appears with timestamp
  4. Generate downstream - Hooks and posts can now be generated

​
Approval bar states

  • Unapproved (yellow warning icon): “This PRF hasn’t been approved yet.”
  • Approved (green checkmark): “Approved on [timestamp] by [user]”
File reference: src/pages/EpisodeDetailPage.tsx:565-577
Approval is a tracking mechanism, not a lock. You can still edit approved PRFs—the approval status simply helps teams coordinate.

​
Revoking approval

If you make significant edits after approval:
  1. Click “Revoke Approval” in the approval bar
  2. Make your edits
  3. Re-approve when ready
This signals to the team that the PRF is back in draft state. File reference: src/pages/EpisodeDetailPage.tsx:571-577

​
Customizing PRF generation

Control how PRFs are generated through Settings:

​
AI model selection

Choose your preferred model:
  • Claude 3.5 Sonnet (default) - Best balance of quality and speed
  • GPT-4 Turbo - Alternative with different writing style
  • Claude 3 Opus - Highest quality, slower
Settings path: Settings > AI Model

​
PRF agent prompt

Customize the PRF generation prompt:
  1. Go to Settings > Agents
  2. Select PRF Agent
  3. Edit the System Prompt
  4. Use preset fields like {episodeNumber} and {guestName}
  5. Save changes
File reference: src/pages/EpisodeDetailPage.tsx:130-141
Preset fields allow dynamic prompt customization:
  • {episodeNumber} - Episode number (e.g., “348”)
  • {guestName} - Guest name (e.g., “Mark Baker”)
  • {recentStyles} - Recently used visual styles (for variety)

​
What happens after PRF generation?

Once the PRF is created:
1

Episode status updates

Status changes from “Draft” to “In Progress”.
2

Guest status updates

Linked guest status changes to “In Production”.
3

Hooks generation triggers

The system automatically generates viral hooks from the PRF.
4

Tabs unlock

Posts, Videos, and Visual suggestions become available for generation.
File reference: src/pages/EpisodeDetailPage.tsx:241-254

​
Fact verification

The PRF agent includes automatic fact-checking:

​
What gets verified

  • Statistics and numbers - Cross-checked against transcript
  • Direct quotes - Verified for accuracy
  • Claims and assertions - Confirmed the guest actually said it

​
Verification process

The agent:
  1. Extracts fact-checkable items from generated PRF
  2. Searches the transcript for supporting evidence
  3. Flags unverified or misattributed content
  4. Includes confidence scores
If fact verification finds critical issues, you’ll see a warning notification. Review flagged content before generating downstream assets.
File reference: src/services/ai.ts:27-49

​
Troubleshooting

Possible causes:
  • Transcript too short (< 500 words)
  • AI model API error
  • Network timeout
Solution: Check transcript length, try regenerating, or switch AI models in Settings.
Possible causes:
  • Transcript has major transcription errors
  • Episode lacks clear structure or themes
  • Default prompt doesn’t match content style
Solution:
  • Clean up transcript before regenerating
  • Customize the PRF agent prompt in Settings
  • Manually edit the PRF to strengthen weak sections
Possible causes:
  • Default prompt uses generic language
  • RAG search didn’t find relevant brand examples
Solution:
  • Customize the PRF agent system prompt with YBH brand voice examples
  • Add brand voice guidelines to Pinecone vector database
  • Manually edit the PRF to match brand tone

​
Next steps

Hooks generation

Generate viral hooks from your PRF

Social posts

Create LinkedIn and Instagram content

Visual suggestions

Generate infographics and quote cards

Episode overview

Return to episode workflow overview

Build docs developers (and LLMs) love

Get started for freeTalk to us