blog-backup/PODCAST_SETUP.md

# Podcast Setup Guide

This guide covers setting up and using the podcast feature for the Daily Digest blog.

## Overview

The podcast feature automatically converts Daily Digest blog posts into audio format using Text-to-Speech (TTS) and provides:

- 🎧 **Web Player** - Listen directly on the blog
- 📱 **RSS Feed** - Subscribe in any podcast app (Apple Podcasts, Spotify, etc.)
- 🔄 **Auto-generation** - TTS runs automatically when new posts are created
- 💾 **Audio Storage** - Files stored in Supabase Storage (free tier)

## Architecture

```
┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│  Daily      │────▶│  TTS API    │────▶│  Supabase   │
│  Digest     │     │ (OpenAI/    │     │  Storage    │
│  Post       │     │  Piper)     │     │             │
└─────────────┘     └─────────────┘     └──────┬──────┘
                                               │
                                               ▼
                                        ┌─────────────┐
                                        │  RSS Feed   │
                                        │  (/api/     │
                                        │  podcast/   │
                                        │  rss)       │
                                        └──────┬──────┘
                                               │
                                               ▼
                                        ┌─────────────┐
                                        │  Podcast    │
                                        │  Apps       │
                                        │ (Apple,     │
                                        │  Spotify)   │
                                        └─────────────┘
```

## Quick Start

### 1. Configure Environment Variables

Add to your `.env.local` and `.env.production`:

```bash
# Enable TTS generation
ENABLE_TTS=true

# Choose TTS Provider: "openai" (paid, best quality) or "macsay" (free, macOS only)
TTS_PROVIDER=openai

# For OpenAI TTS (recommended)
OPENAI_API_KEY=sk-your-key-here
TTS_VOICE=alloy  # Options: alloy, echo, fable, onyx, nova, shimmer
```

### 2. Update Database Schema

Run this SQL in your Supabase SQL Editor to add audio columns:

```sql
-- Add audio URL column
ALTER TABLE blog_messages 
ADD COLUMN IF NOT EXISTS audio_url TEXT;

-- Add audio duration column  
ALTER TABLE blog_messages 
ADD COLUMN IF NOT EXISTS audio_duration INTEGER;

-- Create index for faster RSS feed queries
CREATE INDEX IF NOT EXISTS idx_blog_messages_audio 
ON blog_messages(audio_url) 
WHERE audio_url IS NOT NULL;
```

### 3. Create Supabase Storage Bucket

The app will automatically create the `podcast-audio` bucket on first use, or you can create it manually:

1. Go to Supabase Dashboard → Storage
2. Click "New Bucket"
3. Name: `podcast-audio`
4. Check "Public bucket"
5. Click "Create"

### 4. Deploy

```bash
npm run build
vercel --prod
```

### 5. Subscribe to the Podcast

The RSS feed is available at:
```
https://blog-backup-two.vercel.app/api/podcast/rss
```

**Apple Podcasts:**
1. Open Podcasts app
2. Tap Library → Edit → Add a Show by URL
3. Paste the RSS URL

**Spotify:**
1. Go to Spotify for Podcasters
2. Submit RSS feed

**Other Apps:**
Just paste the RSS URL into any podcast app.

## TTS Providers

### Option 1: OpenAI TTS (Recommended)

**Cost:** ~$2-4/month for daily 5-minute episodes

**Pros:**
- Excellent voice quality
- Multiple voices available
- Simple API integration
- Fast processing

**Cons:**
- Paid service
- Requires API key

**Setup:**
```bash
TTS_PROVIDER=openai
OPENAI_API_KEY=sk-your-key-here
TTS_VOICE=alloy  # or echo, fable, onyx, nova, shimmer
```

### Option 2: macOS `say` Command (Free)

**Cost:** $0

**Pros:**
- Free, built into macOS
- No API key needed
- Works offline

**Cons:**
- Lower voice quality
- macOS only
- Limited to macOS deployment

**Setup:**
```bash
TTS_PROVIDER=macsay
TTS_VOICE=Samantha  # or Alex, Victoria, etc.
```

### Option 3: Piper TTS (Free, Local)

**Cost:** $0

**Pros:**
- Free and open source
- High quality neural voices
- Runs locally (privacy)
- No rate limits

**Cons:**
- Requires downloading voice models (~100MB)
- More complex setup
- Requires local execution

**Setup:**
1. Install Piper:
   ```bash
   brew install piper-tts  # or download from GitHub
   ```

2. Download voice model:
   ```bash
   mkdir -p models
   cd models
   wget https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/lessac/medium/en_US-lessac-medium.onnx
   wget https://huggingface.co/rhasspy/piper-voices/resolve/v1.0.0/en/en_US/lessac/medium/en_US-lessac-medium.onnx.json
   ```

3. Configure:
   ```bash
   TTS_PROVIDER=piper
   PIPER_MODEL_PATH=./models/en_US-lessac-medium.onnx
   ```

## Usage

### Automatic Generation

When `ENABLE_TTS=true`, audio is automatically generated when a new digest is posted via the `/api/digest` endpoint.

### Manual Generation

Generate audio for a specific post:
```bash
npm run generate-tts -- <post_id>
```

Generate audio for all posts missing audio:
```bash
npm run generate-tts:all
```

Force regeneration (overwrite existing):
```bash
npm run generate-tts -- <post_id> --force
```

## API Endpoints

### GET /api/podcast/rss

Returns the podcast RSS feed in XML format with iTunes extensions.

**Headers:**
- `Accept: application/xml`

**Response:** RSS 2.0 XML feed

### POST /api/digest (Updated)

Now accepts an optional `generateAudio` parameter:

```json
{
  "date": "2026-02-23",
  "content": "# Daily Digest\n\nToday's news...",
  "tags": ["daily-digest", "ai"],
  "generateAudio": true  // Optional, defaults to true
}
```

## Database Schema

The `blog_messages` table now includes:

| Column | Type | Description |
|--------|------|-------------|
| `audio_url` | TEXT | Public URL to the audio file in Supabase Storage |
| `audio_duration` | INTEGER | Estimated duration in seconds |

## File Structure

```
src/
├── app/
│   ├── api/
│   │   ├── digest/route.ts          # Updated with TTS trigger
│   │   └── podcast/
│   │       └── rss/route.ts         # RSS feed endpoint
│   ├── podcast/
│   │   └── page.tsx                 # Podcast web player page
│   └── page.tsx                     # Updated with audio player
├── lib/
│   ├── tts.ts                       # TTS service abstraction
│   ├── storage.ts                   # Supabase storage helpers
│   └── podcast.ts                   # RSS generation utilities
└── scripts/
    └── generate-tts.ts              # Manual TTS generation script
```

## Cost Analysis

### OpenAI TTS (Daily Digest ~5 min)
- Characters per day: ~4,000
- Cost: $0.015 per 1,000 chars (tts-1)
- Monthly cost: ~$1.80
- HD voice (tts-1-hd): ~$3.60/month

### Supabase Storage
- Free tier: 1 GB storage
- Audio files: ~5 MB per episode
- Monthly storage: ~150 MB (30 episodes)
- Well within free tier

### Total Monthly Cost
- **OpenAI TTS:** ~$2-4/month
- **Supabase Storage:** $0 (free tier)
- **RSS Hosting:** $0 (Next.js API route)
- **Total:** ~$2-4/month

## Troubleshooting

### TTS not generating
1. Check `ENABLE_TTS=true` in environment variables
2. Check `TTS_PROVIDER` is set correctly
3. For OpenAI: Verify `OPENAI_API_KEY` is valid
4. Check Vercel logs for errors

### Audio not playing
1. Check Supabase Storage bucket is public
2. Verify `audio_url` in database is not null
3. Check browser console for CORS errors

### RSS feed not updating
1. RSS is cached for 5 minutes (`max-age=300`)
2. Check that posts have `audio_url` set
3. Verify RSS URL is accessible: `/api/podcast/rss`

## Future Enhancements

- [ ] Background job queue for TTS generation (using Inngest/Upstash)
- [ ] Voice selection per post
- [ ] Chapter markers in audio
- [ ] Transcript generation
- [ ] Podcast analytics

## Resources

- [OpenAI TTS Documentation](https://platform.openai.com/docs/guides/text-to-speech)
- [Piper TTS GitHub](https://github.com/rhasspy/piper)
- [Apple Podcasts RSS Requirements](https://help.apple.com/itc/podcasts_connect/#/itcb54333f1)
- [Podcast RSS 2.0 Spec](https://cyber.harvard.edu/rss/rss.html)