Developer ports OpenMOSS, ThinkSound, and Trellis to ggml for local asset generation

A developer has ported three open-source models—OpenMOSS for text-to-speech, ThinkSound for sound effects, and Trellis.2 for 3D generation—to the GGML framework to create a complete local pipeline for generating game assets.

• OpenMOSS provides full voice cloning and generation capabilities for NPC characters.
• ThinkSound handles sound effect generation, addressing a gap left by voice models.
• Trellis.2 serves as the current state-of-the-art for open-source 3D generation, including background removal.
• All engines support CUDA, Vulkan, and ROCm for broad hardware compatibility.

These tools are now integrated into the Lemonade SDK, enabling cascading model calls such as a full text-to-3D cycle using stablediffusion.cpp.