SongGeneration Studio: Local AI Song Generation with Full Vocal + Instrumental Tracks

⬅️ Back to Tools

SongGeneration Studio

What it isA local AI song generator with full vocal + instrumental output and a clean web UI
PlatformWindows, macOS, Linux (NVIDIA GPU with 10GB+ VRAM)
PriceFree, open source
Linkgithub.com/BazedFrog/SongGeneration-Studio

Most AI music tools are cloud-only: write a prompt, wait for generation, download the result. SongGeneration Studio wraps Tencent’s LeVo model in a local web UI that gives you control over every part of the song (structure, genre, voice, stems), without sending anything to a server.

  1. You build songs section by section. Add intro, verse, chorus, bridge, outro, or instrumental blocks, drag to reorder, type lyrics for each. The model understands song structure and produces vocals that follow your arrangement. A 3-minute pop song with two verses and three choruses takes about 3-6 minutes to generate on a 24GB GPU.

  2. Style control goes beyond a single genre tag. Pick from pop, rock, hip-hop, R&B, electronic, jazz, metal, folk; then layer in mood (happy, sad, energetic, romantic), voice timbre (male or female with adjustable character), instruments (piano, guitar, drums, synths, strings), and BPM. Or upload a reference track and the AI clones its style.

  3. You get three stems per generation: full mix, vocals only, and instrumental only. The vocal isolation is good enough for remixing or karaoke without running a separate stem splitter. Export to FLAC or MP4 video with generated cover art.

  4. The honest caveat: you need a serious GPU. 10GB VRAM minimum, 24GB recommended. The model download is about 15GB. Generation time varies with song length, and very long songs (5+ minutes) can have quality drops in later sections. The vocal quality is impressive for a local model but doesn’t match cloud services like Suno or Udio on complex arrangements.

Worth your time if: you want to experiment with AI song generation locally, need control over song structure and stems, and have the GPU to run it.

Install & first run

The easiest path is via Pinokio:

  1. Open Pinokio
  2. Search for “SongGeneration Studio”
  3. Click Install (handles dependencies and model download automatically)
  4. Click Start: the web UI opens in your browser

Manual install: clone the repo, run pip install -r requirements.txt, then python main.py. Models download on first launch.

Once the UI loads, write your lyrics in the section editor, pick genre/mood/voice in the style panel, and click generate. Progress streams in real time.

Related TMFNK Content

Crepi il lupo!