pik1 / optimization_report.md
memevis's picture
Upload folder using huggingface_hub
70c4776 verified

TTS Model Optimization Report

Overview

  • Optimization Date: 2025-04-14 17:08:24
  • Iterations Performed: 3
  • Total Strategies Tested: 32
  • Best Overall Score: 0.8708

Performance Metrics

  • Mel Cepstral Distortion: 0.9590
  • Word Error Rate: 0.0548
  • Naturalness: 0.9246
  • Intelligibility: 0.9577
  • Speaker Similarity: 0.9453
  • Prosody: 0.9438
  • Overall Quality: 0.9845

Optimization Insights

Most Effective Parameter Settings

  • attention_scale: 1.2800
  • output_scale: 1.3800
  • projection_scale: 1.6200
  • encoder_scale: 1.6602
  • decoder_scale: 1.2200
  • base_enhancement: 0.0029
  • importance_factor: 1.5999

Optimization Journey

Iteration 1: Best strategy 'intelligibility_focus_var1.5' - Score 0.8693

Iteration 2: Best strategy 'intelligibility_focus_var1.5_iter2_encoder_boost' - Score 0.8701

Iteration 3: Best strategy 'intelligibility_focus_var1.5_iter2_encoder_boost_iter3_noise_scale_1.2' - Score 0.8708