Skip to content

3.1.0

Choose a tag to compare

View all tags
@CoffeeMethod CoffeeMethod released this 14 Feb 00:21
· 89 commits to main since this release
3.1.0
df4af1b

Release Notes - v3.1.0

Major New Features

JIT (Just-In-Time) Generation

Introducing real-time audio streaming! You no longer have to wait for a full document to be processed before you start listening.

  • Real-time Playback: Audio begins playing as soon as the first segments are ready.
  • Smart Buffering: The engine maintains a 2-minute buffer ahead of playback to ensure a smooth, uninterrupted experience.
  • Interruptible: Stop at any time; the system will automatically save what was played and export the remaining text to a _remaining.txt file so you can resume later.

Integrated Audio FX Pipeline

Powered by Spotify's pedalboard library, you can now apply professional-grade audio effects directly within the GUI.

  • Effects included: Reverb (Room/Hall simulation), Compressor (for consistent volume), and EQ (High/Low Shelf filters).
  • FX Presets: Save your favorite effect chains and apply them globally or per-segment via script.

Pronunciation Lexicon

Take control over how the AI speaks.

  • Custom Overrides: Define a "Find & Replace" dictionary for specialized terms, acronyms, or names that the base model might mispronounce.
  • Persistent: Your lexicon is saved to your configuration and applied automatically to all generations.

Advanced Voice Mixing

Unlock a near-infinite variety of voices.

  • Custom Voice Tab: Mix any two standard voices with precise ratio and operation control to create unique voices.
  • Save & Reuse: Export your custom mixes as .pt files to use them just like standard voices.

Enhancements & Fixes

  • Scripted Multi-Speaker & FX: Added support for the advanced syntax [Speaker_Preset:FX_Preset]: Your text here.
  • Intelligent Caching: Implemented a SHA-256 hashing system. If you generate the same text with the same settings twice, the GUI will instantly pull from the cache instead of re-generating.
  • Windows run.bat: A new one-click startup script that automatically handles virtual environment creation and dependency updates.
  • Performance: Optimized the dual-loop architecture for JIT mode to minimize CPU overhead during simultaneous generation and playback.
  • Expanded Language Support: Full integration for Spanish, French, Italian, Portuguese, Japanese, and Chinese.

Requirements Update

This version introduces new dependencies:

  • pedalboard (for Audio FX)
  • winsound (standard on Windows, used for JIT playback)

To update, simply run run.bat or execute:

pip install -r requirements.txt
Assets 2
Loading