Three.js live scene / mode: generate
Build songs, split stems, train voices.
A full AI music command desk with prompt-to-song, legal voice modeling, stem mixing, and export-ready production flow.
TJ MUSIC RESEARCH STUDIO
Turn your idea into a real pop-song pipeline
Suno/Mureka-style workflow, but routed to our open-source engine lab. The old 16/32-bar local demo path is blocked here until a real model returns audio.
Reference Feed
Study, prompt, then generate through engines
For You
Made with Studio
Current readiness
Product boundary
Auth + credits
register, login, plan upgrade, credit debit
Lyrics + song
structured lyrics and prompts; MP3/WAV only after a real model returns audio
Audio split
upload path ready
Voice consent
private model with authorization gate
Studio
mute, solo, volume, download, export
Production radar
Open-source + provider chain
Real song output is blocked in mock mode. The studio can prepare prompts and engine routing, but will not fake finished audio.
primary full-song generation with vocals, lyrics, covers, remix, repaint and style LoRA
local venv/API verified; model weights pending
lyrics and tag driven complete songs with multilingual structure
repo cloned; inference wrapper pending
fast full-song generation with lyric alignment and vocal/accompaniment integration
download blocked by GitHub TLS; needs retry/mirror
long songs, lyrics-to-song and multilingual vocal research
repo cloned; weight and VRAM plan pending
BGM, instrumental, sound-effect layers and audio inpainting
not installed locally yet
Jobs
WebSocket-style progress feed
No jobs yet.