• Use fork of espeak-ng for punctuation preservation and sentence boundary detection
• Support raw text phonemes with piper-phonemize
• Support Arabic diacritization with libtashkeel (model included)
• Extend default phoneme set to 256 for future expansion (use these pretrained checkpoints)
• New command-line options (--silence_seconds, --espeak_data, --tashkeel_model, --debug)
• Merge code into piper.hpp and piper.cpp for use as a library
• Update to onnxruntime 1.14.1 (from piper-phonemize)
• Adding logging with spdlog