k5 / optimization_report.md
memevis's picture
Upload folder using huggingface_hub
ebd1fa6 verified

TTS Model Optimization Report

Overview

  • Optimization Date: 2025-05-17 01:42:50
  • Iterations Performed: 3
  • Total Strategies Tested: 32
  • Best Overall Score: 0.8702

Performance Metrics

  • Mel Cepstral Distortion: 0.9042
  • Word Error Rate: 0.0713
  • Naturalness: 0.9458
  • Intelligibility: 0.9393
  • Speaker Similarity: 0.9504
  • Prosody: 0.9655
  • Overall Quality: 0.9593

Optimization Insights

Most Effective Parameter Settings

  • attention_scale: 1.3000
  • output_scale: 1.4000
  • projection_scale: 1.6000
  • encoder_scale: 1.5000
  • decoder_scale: 1.2000
  • base_enhancement: 0.0015
  • importance_factor: 1.5599

Optimization Journey

Iteration 1: Best strategy 'intelligibility_focus_var0.75' - Score 0.8648

Iteration 2: Best strategy 'intelligibility_focus_var0.75_iter2_noise_scale_1.2' - Score 0.8702

Iteration 3: Best strategy 'intelligibility_focus_var0.75_iter2_noise_scale_1.2_iter3_base_enhancement_1.2' - Score 0.8666