The ComfyUI-FS_Audio_Suite node set has been released for music generation

Developers introduced ComfyUI-FS_Audio_Suite — a modular set of tools for generating audio, integrated into the ComfyUI ecosystem. The package includes six specialized nodes designed for working with audio data and training custom models.

The main functionality of the solution includes a system for converting hums into full-fledged musical compositions (hum-to-song). Users can upload their own LoRA adapters to customize generation for specific styles or instruments. The package also includes a built-in dataset constructor that allows preparing data for training planners (planner LoRA) and decoders.

The tool is focused on creating custom audio models directly in the ComfyUI interface. The solution's architecture supports flexible customization of generation parameters, allowing for the automation of creating musical fragments based on user audio inputs. The project is distributed with open source code.

Implementing this set of nodes simplifies pipelines for authors working with generative audio. Previously, similar tasks required the use of external scripts or separate software environments, but now all stages — from data preparation to inference — are available within the single ComfyUI workspace.