mtmd : add Soprano audio generation

[hans00]

Overview

Support https://huggingface.co/ekwek/Soprano-1.1-80M

Usage

# Convert Qwen3 backbone
python convert_hf_to_gguf.py /path/to/Soprano-1.1-80M \
    --outfile soprano.gguf \
    --outtype f16

# Convert Soprano decoder
python convert_hf_to_gguf.py /path/to/Soprano-1.1-80M \
    --mmproj \
    --mmproj-architecture SopranoModel \
    --outfile mmproj-soprano.gguf \
    --outtype f16

# Run TTS
llama-tts \
    -m soprano.gguf \
    -mm mmproj-soprano.gguf \
    -p "Hello world! This is a test of speech synthesis." \
    --temp 0 \
    -n 512 \
    -ngl 0 \
    --output out.wav

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: Codex