ck9 / optimization_report.md
memevis's picture
Upload folder using huggingface_hub
98d8463 verified

TTS Model Optimization Report

Overview

  • Optimization Date: 2025-05-20 04:21:17
  • Iterations Performed: 3
  • Total Strategies Tested: 34
  • Best Overall Score: 0.8735

Performance Metrics

  • Mel Cepstral Distortion: 0.9304
  • Word Error Rate: 0.0687
  • Naturalness: 0.9448
  • Intelligibility: 0.9564
  • Speaker Similarity: 0.9710
  • Prosody: 0.9451
  • Overall Quality: 0.9791

Optimization Insights

Most Effective Parameter Settings

  • attention_scale: 1.6400
  • output_scale: 1.3400
  • projection_scale: 1.4800
  • encoder_scale: 1.1600
  • decoder_scale: 1.5400
  • base_enhancement: 0.0017
  • importance_factor: 1.7083

Optimization Journey

Iteration 1: Best strategy 'attention_decoder_focus' - Score 0.8735

Iteration 2: Best strategy 'attention_decoder_focus_iter2_base_enhancement_1.2' - Score 0.8697

Iteration 3: Best strategy 'attention_decoder_focus_iter3_noise_scale_1.2' - Score 0.8657