StudioKoboldCpp v1.124 released with YuE2 music generation and new TTS and image model support
Original title: Koboldcpp v1.124 released
KoboldCpp v1.124 (aesthetic edition) is released, integrating a subset of models from audio.cpp to greatly enhance built-in Music Gen and TTS Gen. Music Gen adds support for YuE2 with three templates (no LLM, small 4B LLM for lyrics, big LLM for large GPUs); TTS adds 17 ported models including BreezeTTS 2, Ming 3.0, Qwen3-TTS, VoxCPM2, Chatterbox and CosyVoice 3; the sd.cpp sync brings MOSS Image, Fibo 2 Edit and Flux 3 Klein image generation. AI Horde now supports Chat Completions workers with jinja, reasoning, tool calls, streaming and vision; new additions include a /mcp wizard for custom MCP endpoints, a /paste mode, CUDA graphs for all architectures, --paralleltotalcontext for the shared KV pool, improved Gemma4 SWA handling and WebSearch, and upstream support for Nemotron 3.5 Bolt 7…
The post lists the full changelog for v1.124, letting local deployment users evaluate whether the new music, TTS and image models and backend fixes warrant an upgrade.
The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.