Files
novafarma/docs/BETTER_VOICE_OPTIONS.md
2026-01-20 01:05:17 +01:00

79 lines
2.1 KiB
Markdown

# 🎤 BETTER VOICE OPTIONS - Testing Guide
## Current Issue:
Voices sound too robotic/AI-generated
## Solution:
Test multiple Edge-TTS voices to find most natural sounding
---
## 🎙️ VOICE OPTIONS TO TEST
### **KAI (Young Male, 14 years old)**
**Current:** `en-US-GuyNeural` (energetic but robotic)
**Better alternatives:**
1. `en-US-ChristopherNeural` - Young, warm, natural
2. `en-US-EricNeural` - Teen friendly, less robotic
3. `en-US-RogerNeural` - Mature teen voice
4. `en-GB-RyanNeural` - UK teen, authentic
**Best choice:** `en-US-ChristopherNeural` (most natural for teen)
---
### **ANA (Young Female, 14 years old - twin)**
**Current:** `en-US-JennyNeural` (warm but AI-ish)
**Better alternatives:**
1. `en-US-AriaNeural` - Natural, expressive
2. `en-US-SaraNeural` - Youthful, authentic
3. `en-GB-SoniaNeural` - UK accent, warm
4. `en-US-MichelleNeural` - Soft, emotional
**Best choice:** `en-US-AriaNeural` (most expressive/natural)
---
### **GRONK (Deep UK voice)**
**Current:** `en-GB-RyanNeural` (good!)
**Keep or try:**
1. `en-GB-ThomasNeural` - Deeper, gruffer
2. `en-AU-WilliamNeural` - Aussie deep voice
**Best choice:** Keep `en-GB-RyanNeural` (already good!)
---
## 🎯 GENERATE TEST SAMPLES
```bash
cd /Users/davidkotnik/repos/novafarma/assets/audio/voiceover
# Test Kai voices
python3 -m edge_tts --text "It all started with family. With colors. With hope." --voice en-US-ChristopherNeural --write-media test_kai_christopher.mp3
python3 -m edge_tts --text "It all started with family. With colors. With hope." --voice en-US-EricNeural --write-media test_kai_eric.mp3
# Test Ana voices
python3 -m edge_tts --text "We were unstoppable. We were immortal." --voice en-US-AriaNeural --write-media test_ana_aria.mp3
python3 -m edge_tts --text "We were unstoppable. We were immortal." --voice en-US-SaraNeural --write-media test_ana_sara.mp3
```
---
## ✅ FINAL RECOMMENDATION
**Use these voices:**
- **Kai:** `en-US-ChristopherNeural` (natural teen voice)
- **Ana:** `en-US-AriaNeural` (expressive, emotional)
- **Gronk:** `en-GB-RyanNeural` (keep - already good!)
**Regenerate all 21 files with better voices!**