Description

Two Konata Voice Models! 🎵 Konata’s tone shifts a lot (+9 to +15/16 for male, -3 to +3 for female), so these models are tuned to capture that perfectly. Small Model: Quick & light, great for real-time voice changes! Trained on 10 minutes of data. Mid Model: Larger, slower, but much higher quality. Best for achieving a better result! Trained on 40 minutes of data. I originally planned a High Model trained on every voice line from the anime, but I stopped around 10-12 episodes (which ended up giving us the Mid Model). Realized the extra data wouldn’t make a huge difference in quality, so the Mid Model is solid enough for top-tier results!