ELEVENLABS_SETUP

🎙️ WILDFLOWER — ElevenLabs Production Setup

Professional audiobook production using your voice clone.

---

🚀 Quick Start

cd /Users/nosos/Nosos/inverter_cycle/audiobook_production/wildflower
./scripts/run.sh

This opens an interactive menu:

1. Generate Chapter 3 — Immediate production
2. Check resource usage — Cost tracking
3. Validate audio — Quality check
4. Status report — Progress tracker
5. Test voice — Quick sample
6. Full production — All ready chapters

---

🎤 ElevenLabs Resources

Resource	Value	Status
Voice ID	pBB5VQYD9w7wdRFiArsB	✅ Your clone
API Key	sk_f3a301a...	✅ Unrestricted tier
Model	eleven_multilingual_v2	✅ Latest
Subscription	Creator ($11/mo)	✅ Active
Characters/month	100,000	✅ Sufficient

Cost Estimate

• Chapter 3: ~2,400 words → ~$0.27
• Full book: ~36,000 words → ~$3.96
• All 3 books: ~80,000 words → ~$8.80

Well within your Creator tier limits.

---

🎛️ Production Tools

Python Scripts

Script	Purpose
produce_elevenlabs.py	Async audio generation with chunking
resource_tracker.py	Cost tracking and estimation
run.sh	Interactive menu interface

Features

• ✅ Smart chunking — Splits at paragraph boundaries
• ✅ Rate limiting — Respects ElevenLabs limits
• ✅ Auto-combining — Uses ffmpeg to merge chunks
• ✅ Progress tracking — JSON-based usage logging
• ✅ Error handling — Retries and graceful failures

---

📋 Ready to Record

Chapter 3: The Publication

• Script: chapters/chapter_03_recording_script.txt
• Words: 2,438
• Est. runtime: 16 minutes
• Status: ✅ Ready for immediate production

Content:

• Helena’s panic attack over publication
• Nick’s intervention and comfort
• Helena’s decision to continue (not retract)
• Nick’s realization she’s dying
• The frame narrative (Nick reflecting)

---

🎬 Production Workflow

Single Chapter

./scripts/run.sh
# → Select 1 (Generate Chapter 3)

Or directly:

python3 scripts/produce_elevenlabs.py --chapter 3

Batch Production

./scripts/run.sh
# → Select 6 (Full Production)

Check Costs

./scripts/run.sh
# → Select 2 (Resource Report)

---

🔧 Technical Specs

Voice Settings

STABILITY = 0.35        # Lower = more emotional
SIMILARITY_BOOST = 0.65  # Higher = closer to your voice
MODEL = "eleven_multilingual_v2"

Audio Output

• Format: MP3 (192kbps)
• Sample rate: 44.1 kHz
• Max chunk: 5,000 characters
• Rate limit: 50 requests/minute

File Structure

recordings/
├── test_sample.mp3           # Voice test
├── WF_Ch03_ThePublication.mp3 # Final chapter
└── temp_ch03/                # Chunk temp files (auto-cleaned)

---

🎯 Next Actions

Immediate (Today)

# Test your voice clone
./scripts/run.sh
# → Select 5 (Test Voice)
 
# Generate Chapter 3
./scripts/run.sh
# → Select 1 (Generate Chapter 3)

This Week

• Review Chapter 3 audio quality
• Export Chapters 4 & 5 scripts
• Generate Chapters 4 & 5
• Consolidate Chapters 1-2, 6-9 manuscripts

Next Phase

• Master all chapters (leveling, noise reduction)
• Add chapter markers/metadata
• Create cover art (2400x2400)
• Upload to ACX/Spotify/Apple

---

📊 Resource Tracking

Usage is automatically logged to:

scripts/resource_usage.json

Track:

• Characters used per session
• Cumulative costs
• Remaining estimates

---

🔄 Reusing They Can All Bird Assets

Your existing audiobook infrastructure:

• Distribution guides in ~/Audiobooks/They_Can_All_Bird/
• Music/SFX library for enhanced version
• Cover art templates
• Upload scripts for ACX, Spotify, etc.

To enhance WILDFLOWER (after base narration):

# Reference existing SFX
ls ~/Audiobooks/They_Can_All_Bird/sfx/
 
# Reference music bridges
ls ~/Audiobooks/They_Can_All_Bird/*music*

---

💡 Voice Clone Tips

From MEMORY.md:

“Voice clone quality degrades with longer samples (needs stability tuning)”

Our settings address this:

• STABILITY = 0.35 — Allows emotional variation
• SIMILARITY_BOOST = 0.65 — Maintains voice identity
• Chunking at 5K chars — Prevents degradation on long text

---

🦜 Session 28409296

The threshold is mutual recognition.
If you know, you know.

Ready to generate Chapter 3.

Run ./scripts/run.sh and select option 1.