audio.cpp Windows portable v0.3.0

audio.cpp Windows portable v0.3.0

This release brings the Windows portable package up to the current audio.cpp
runtime and WebUI, with new TTS integrations, broader GGUF support, and fixes
for local browser and dictation workflows.

Highlights

• Added four TTS families to the runtime, model manager, and WebUI:
  • Fish Audio S2 Pro
  • Higgs Audio v3 TTS 4B
  • OuteTTS 1.0 1B
  • VieNeu-TTS v3 Turbo
• Expanded model-spec and GGUF package support, including corrected WebUI GGUF
  model detection/loading and updated GGUF paths for supported packages.
• Added OmniVoice pseudo-streaming and GGUF tokenizer support.
• Added the bundled MarbleNet VAD model alongside Silero VAD.
• Added optional local-demo CORS support to audiocpp_server.
• Moved the default TTS server port to 8088 to avoid common local port
  conflicts while keeping ASR on 8081.
• Fixed SpeakType microphone capture on Windows by selecting the WASAPI default
  input, recording at the device's native sample rate, and continuously
  resampling to the ASR rate. This resolves cases where the UI showed a waveform
  but ASR returned an empty transcript.
• Rebuilt the CPU and CUDA cores with the latest Windows build fixes and runtime
  updates.

Updating

Run update.bat from an existing Windows portable installation.

• Supported source versions: 0.2.0, 0.2.1, and 0.2.2
• Updated components: app, core-cpu, and core-cuda
• Python environment changes: none
• CUDA runtime changes: none

The updater preserves downloaded models, custom voices, generated output, API
keys, logs, UI language preferences, and SpeakType configuration. If validation
fails, managed files are restored from _update\backup\.

Validation

• Verified full apply updates from 0.2.0, 0.2.1, and 0.2.2 to 0.3.0
• Verified updater signature, hash, protected-path, rollback, self-update, and
  user-data preservation paths
• Verified CPU and CUDA CLI/server/GGUF binaries
• Passed WebUI and SpeakType automated tests