Skip to content

v2.0.1 - UI Overhaul & Community Improvements

Choose a tag to compare

View all tags
@ReSerendipity ReSerendipity released this 23 Jul 13:28
· 552 commits to main since this release
v2.0.1
538cf84

Highlights

  • Dual-engine architecture: VoxCPM2 + IndexTTS2 with runtime switching
  • Emotion tag system: 30+ emotion/style tags for fine-grained voice control
  • Neural watermarking: Spread-spectrum audio watermarking for ownership protection
  • vLLM backend: Optional acceleration for high-throughput inference
  • CLI batch processing: JSON/CSV input with progress reporting and ETA
  • Enhanced documentation: Model download guide, examples, contribution guide
  • GitHub community: Release notes config, issue templates, Good First Issues

New Features

  • emotion_tags.py — 30+ emotion/style tags inspired by Fish Speech
  • watermark.py — Neural watermarking inspired by Chatterbox Perth
  • vllm_backend.py — Optional vLLM integration with auto-fallback
  • cli.py — Batch processing with JSON/CSV input support
  • examples/ — API usage examples (clone, design, batch)

Documentation

  • Fixed model download guide with correct repo IDs
  • Added "Why TTS MultiModel?" section to README
  • Enhanced CONTRIBUTING.md with dev setup and Good First Issues

Design Documents

  • Model Tiering Plan (Turbo/Nano/Standard)
  • Training Toolchain Plan
  • RL Alignment Plan (GRPO)
  • TensorRT Integration Plan
  • Dialect Support Plan
  • Multi-Speaker Plan
  • Edge Deployment Plan
  • OpenAI Compatible API Plan

Full Changelog: https://github.com/ReSerendipity/TTS_MultiModel/blob/master/CHANGELOG.md

Assets 2
Loading