RDF STORYTELLERS v0.4c - SELECTED NEURAL VOICES

LOCKED VOICES
=============
PlayDrone
  Voice : Microsoft Ava Neural (en-US-AvaNeural)
  Style : Teasing Playful
  Rate  : +3%
  Pitch : natural (+0 Hz)

Mistress
  Voice : Microsoft Michelle Neural (en-US-MichelleNeural)
  Style : Soft Commanding
  Rate  : -7%
  Pitch : almost natural (-1 Hz)

These are the exact two audition directions selected for the mod.

WHAT CHANGED
============
- The old Windows SAPI/Desktop generator is no longer used.
- PlayDrone and Mistress now use modern neural female voices.
- All matching lines in VoiceLines.tsv are generated with these selected voices.
- Generated game audio is converted to mono 24 kHz PCM WAV for RimWorld.
- No vocoder, metallic filter, bitcrush, echo, or heavy pitch shifting.
- Cassandra, Phoebe and Randy remain text-only.
- Commentary logic, event detection, event weighting and RDF Events integration are unchanged.
- Robotics is not touched.

BUILD
=====
1. Put the mod folder into RimWorld/Mods.
2. Run PreviewSelectedVoices.bat if you want to hear the locked voices again.
3. Run Build.bat.

The first voice generation needs:
- Python 3
- Internet access

Build.bat creates a local _voice_env and installs:
- edge-tts
- imageio-ffmpeg

No API key is needed.

After the voice pack has been generated, the WAV files remain in:
  Sounds/StorytellerVoices/

The normal game runtime does NOT need Python, internet access, edge-tts or ffmpeg.

EXPECTED LOG LINE
=================
[RDF Storytellers] v0.4c selected-neural-voice commentary installed.

REFERENCE CLIPS
===============
The exact selected audition MP3 files are included under:
  Docs/SelectedVoiceReferences/
