AI Assistant

preroll.io includes a built-in AI assistant that handles the repetitive content work of podcast production — transcription, show notes, episode descriptions, social posts, and title suggestions. It also provides a conversational chat interface where you can ask questions about your data or take actions hands-free.

The AI features are available as a metered add-on, separate from your plan tier. Credits are shared across all AI operations.

AI Chat

The AI chat is a conversational assistant that understands your preroll.io data. Open it from any page in the app to ask questions, look things up, or perform actions without navigating away from what you're doing.

Opening the Chat

• Click the sparkle icon in the bottom-right corner of any page
• Press Cmd+K (Mac) or Ctrl+K (Windows/Linux)
• Press Escape to close

The chat panel slides in from the right side and persists across page navigation — you can keep a conversation going while moving between episodes, shows, and clients.

Context Awareness

The assistant automatically knows where you are in the app:

• On an episode page — "Generate show notes for this episode" works without specifying which episode
• On a show page — "What episodes are due this week?" scopes to that show
• On a client page — "Add a meeting note" knows which client you mean
• On the dashboard — General questions about your workspace

What It Can Do

Look things up (instant):

• List episodes by status, deadline, or show
• Check pipeline progress across shows
• View upcoming deadlines
• Check AI credit balance and usage
• Summarize recent activity
• Search across clients, shows, and episodes

Take actions (with confirmation):

• Create clients, shows, and episodes
• Move episodes between pipeline stages
• Update episode details (title, description, notes, deadline)
• Share files with clients
• Add meeting notes
• Trigger transcription or content generation
• Create and manage tags

Write operations always show a Confirm / Cancel card before executing — the assistant will never modify your data without your explicit approval.

Not available via chat:

• Deleting anything (too destructive)
• Billing and subscription changes
• Org settings or team management
• OAuth integration setup
• Transistor publishing (use the dedicated publish UI)

Session Management

Conversations are saved automatically. You can:

• Start a new conversation with the + button
• Switch between recent sessions from the dropdown in the chat header
• Sessions are auto-titled based on your first message

Credits

Each chat message costs 1 AI credit. If the assistant triggers transcription or content generation on your behalf, those operations consume additional credits at their normal rates.

Transcription

preroll.io transcribes episode audio using Deepgram Nova-2, producing speaker-diarized transcripts that serve as the foundation for all AI content generation.

How to Transcribe

1. Link an audio or video file to your episode (via upload, Frame.io, Google Drive, or Vimeo)
2. Open the episode detail page
3. Click Transcribe in the AI section (or ask the chat assistant)
4. Transcription runs in the background — the page updates automatically when complete

Transcription typically takes 30-60 seconds for a 1-hour episode.

Auto-Transcription

You can configure shows to automatically transcribe new audio:

1. Go to Show Settings → AI
2. Enable Auto-transcribe

When enabled, transcription starts automatically whenever:

• A file is uploaded or linked to an episode
• A Frame.io, Google Drive, or Vimeo webhook reports a file is ready

Transcript Features

• Speaker diarization — Identifies distinct speakers
• Word-level timestamps — Segments are timecoded for reference
• Searchable — Full-text search within the transcript
• Copyable — One-click copy of the full transcript text

Credit Cost

Transcription costs 1 credit per minute of audio (rounded up).

Content Generation

Once an episode has a transcript, you can generate production and promotion content from it.

Content Types

Generating Content

From the episode detail page:

1. Scroll to the AI section
2. Click Run AI Pipeline to generate all enabled types, or click individual type buttons to generate one at a time
3. Generated content appears inline with options to Copy, Edit, Apply to Episode, or Regenerate

From the chat assistant:

"Generate show notes for this episode"

Per-Show Configuration

Each show can have its own AI settings (Show Settings → AI):

These settings apply to both manual and automatic generation.

Applying Generated Content

For show notes and descriptions, you can Apply to Episode to write the generated text directly into the episode's fields. A confirmation dialog prevents accidental overwrites.

For social posts, use the Copy button on individual posts or Copy All to bundle all promotion content.

Title suggestions are presented as individual selectable options — click Copy on whichever title you prefer.

Editing Before Applying

Click Edit on any generated content to modify it inline. After editing, you can Save & Apply to write your modified version to the episode, or Copy edited to grab it for use elsewhere.

Generation History

Previous versions of generated content are preserved. Click "N previous" below any generation to expand the history, compare versions, and copy older outputs.

Batch Processing

From a show page, click Run AI to batch-process up to 5 episodes that have audio but no transcript. This queues transcription and auto-generation for each eligible episode.

AI Pipeline

The AI pipeline is the automated flow from audio file to finished content:

Audio file linked → Transcribe → Generate enabled content types

Automatic Triggers

The pipeline runs automatically when:

1. File upload — An audio/video file is linked to an episode via file_references
2. Integration webhook — Frame.io, Google Drive, or Vimeo reports a file is ready
3. Manual trigger — You click "Run AI Pipeline" or ask the chat assistant

Pipeline Progress

The AI section shows a step-by-step progress indicator:

• Transcription — In progress / Complete
• Show Notes — Generating / Complete
• Description — Generating / Complete
• ...additional enabled types

A notification banner appears when the pipeline completes, and the panel auto-expands if it was collapsed.

Credits and Billing

AI features use a shared credit pool, separate from your plan subscription.

How Credits Work

• Monthly allowance — Pro plans include 200 credits/month, Studio includes 500
• Purchased credits — Buy additional credit packs that never expire
• Credits are consumed in order: monthly first, then purchased
• Unused monthly credits do not roll over

Credit Packs

Purchase from Settings → AI.

Cost Reference

Monitoring Usage

• The AI section on episode pages shows your current balance
• Settings → AI shows full usage history
• The chat assistant can tell you your balance: "How many credits do I have left?"
• Low-credit warnings appear when your balance is running low

Plan Requirements

Free-plan users see the AI chat button but receive an upgrade prompt when attempting to use it.

Self-Hosted Mode

Self-hosted deployments can use their own API keys instead of credits:

1. Set PREROLL_SELF_HOSTED=true in your environment
2. Go to Settings → AI
3. Enter your own Deepgram API key (for transcription) and Anthropic API key (for generation and chat)

With BYOK (Bring Your Own Key), no credits are consumed — usage is billed directly by the respective providers.