ck7 / optimization_report.md
memevis's picture
Upload folder using huggingface_hub
b09948e verified

TTS Model Optimization Report

Overview

  • Optimization Date: 2025-05-19 08:53:54
  • Iterations Performed: 3
  • Total Strategies Tested: 34
  • Best Overall Score: 0.8739

Performance Metrics

  • Mel Cepstral Distortion: 0.9063
  • Word Error Rate: 0.0502
  • Naturalness: 0.9443
  • Intelligibility: 0.9465
  • Speaker Similarity: 0.9630
  • Prosody: 0.8916
  • Overall Quality: 0.9672

Optimization Insights

Most Effective Parameter Settings

  • attention_scale: 1.8803
  • output_scale: 1.5698
  • projection_scale: 1.4398
  • encoder_scale: 1.0995
  • decoder_scale: 1.7361
  • base_enhancement: 0.0016
  • importance_factor: 1.6602

Optimization Journey

Iteration 1: Best strategy 'prosody_expression_var0.75' - Score 0.8739

Iteration 2: Best strategy 'prosody_expression_var0.75_iter2_noise_scale_0.8' - Score 0.8581

Iteration 3: Best strategy 'prosody_expression_var0.75_iter3_base_enhancement_0.8' - Score 0.8673