HOT-Step-CPP Ignites Offline AI Music Creation On Your Own GPU

HOT-Step-CPP is a new release that delivers a complete, locally-run AI music studio with a polished graphical interface. It wraps the acestep.cpp inference engine, letting you describe a song with text and lyrics to generate stereo 48kHz audio without any cloud services or subscriptions. The project arrives as a portable package where you can just extract the files, launch a script, and start creating.
Developer scragnog built this as a feature-rich UI extension for the GGML-powered acestep.cpp engine. The tool bundles over 100 extra capabilities ranging from advanced audio post-processing to a full lyrics workstation and stem separation. Everything runs on your own hardware using CUDA, Vulkan, Metal, or CPU backends.
Full creative control from idea to finished track
- AI-powered song generation from genre and subject.
- Manual mode for lyrics, BPM, key, and structure.
- Lyric studio with 7 LLM providers and editors.
- Cover studio with automatic BPM and key detection.
- 4-stage neural stem separation and interactive mixer.
- VST3 plugin host for custom effects and processing.
- Region-based audio repaint to fix specific sections.
- Side-by-side A/B comparison of generated tracks.
This tool suits privacy-focused musicians, small production studios, and anyone with a capable consumer GPU who wants to stay offline. You can go from a genre idea to a fully mastered song with lyrics, stems, and alternate covers without ever touching a cloud API. The built-in model manager lets you download everything you need right from the app.
What’s running under the hood
The entire stack combines a C++ engine using CUDA and GGML, a Node.js server, and a React-based frontend. Pre-built releases exist for Windows, Linux, and Apple Silicon with no installation required — just extract and run. The plugin system is extensible with Lua scripts, so new solvers and postprocessors can appear in the UI without rebuilding the C++ code.
"Describe a song with a text caption and lyrics, and get stereo 48kHz audio generated entirely on your local hardware. No cloud, no API keys, no subscriptions." — Source: GitHub