coolcoder613/omnivoice-latin-GGUF-1

OmniVoice Latin (GGUF)

GGUF conversion of k2-fsa/OmniVoice with a Latin LoRA merged in (restored classical pronunciation, ~22 h of macronized Latin), for omnivoice.cpp. LoRA adapter version: https://huggingface.co/coolcoder613/omnivoice-latin-lora-1.

• omnivoice-base-Q8_0.gguf recommended; omnivoice-base-BF16.gguf near-lossless
• omnivoice-tokenizer-F32.gguf audio codec (unchanged from upstream)

Latin is not in omnivoice.cpp's language table. Before building, add this line to src/lang-map.h after the "latgalian" entry:

{ "latin",   "la"  },

Then:

echo "Gallia est omnis dīvīsa in partēs trēs." | ./build/omnivoice-tts \
    --model omnivoice-base-Q8_0.gguf --codec omnivoice-tokenizer-F32.gguf \
    --lang Latin -o out.wav