YuE2 Studio is a Windows desktop app by Ilya Timonin (Nerual Dreming) for making full songs on your own GPU with YuE2, the open song model from M-A-P that writes a score before it sings. You describe a style and write lyrics, the model composes a melody with chords as sheet music in ABC notation and then performs it with vocals. The score comes back with the track, so you can edit the notes, tempo or key and render the same composition again with a different sound. The studio is Rust and React in a Tauri window, built on yue2.cpp, Serveurperso's native C++/CUDA port of YuE2, so nothing in the runtime path is Python. It ships as an installer with auto-update or a portable folder and runs offline. Songs run up to six minutes, and the developer says a 3:38 song renders in about 46 seconds on an RTX 4090 with the Q8_0 set. Covers use SheetSage2 to write a recording's melody as a score that YuE2 then sings in a new style, and every track keeps its request and audio codes for exact replay. The app also bundles 110 examples, a writing assistant (a local Gemma model, OpenRouter or a connected agent), word-level karaoke lyrics, six-stem separation with HT-Demucs, audio-to-MIDI, a MIDI editor, LoRA loading and training, audio processing, music videos and an MCP server with 162 tools. It needs Windows 10/11 x64, a GPU with 6 GB of VRAM or more and 4-10 GB of disk for one model set. NVIDIA through CUDA is the tested path; AMD and Intel through Vulkan are experimental, and the CPU fallback is many times slower. Training a LoRA needs an NVIDIA RTX 30-series card or newer with 11 GB of VRAM. The studio is MIT, but the YuE2-3B and VAE weights are CC BY-NC 4.0 with an individual-creator permission, and companies that want to use them commercially need a license from the authors; SheetSage2, used for covers, is CC BY-NC 4.0 without that permission.