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 routed to our open-source engine lab: style choice, PromptBrain expansion, time-coded structure, anti-repeat changes, final waveform render, then download.
SCORE ROBOT
Score Robot inside the workstation
Scene brief to scoring plan, then into this same Create Song pipeline. The standalone robot remains at /composer.
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
A real local song engine is available. Full hosted production still needs the infrastructure shown below.
primary full-song generation with vocals, lyrics, covers, remix, repaint and style LoRA
随工作站附带的默认出歌引擎 / Default song engine included with the workstation.
lyrics and tag driven complete songs with multilingual structure
可选引擎:歌词与标签驱动的整首歌 / Optional engine: lyrics- and tag-driven full songs.
fast full-song generation with lyric alignment and vocal/accompaniment integration
可选引擎:快速成曲与歌词对齐 / Optional engine: fast full-song generation with lyric alignment.
long songs, lyrics-to-song and multilingual vocal research
可选引擎:长篇幅歌曲,显存需求较高 / Optional engine for long-form songs; needs more VRAM.
BGM, instrumental, sound-effect layers and audio inpainting
可选引擎:纯音乐、BGM 与音效 / Optional engine for instrumentals, BGM and sound effects.
Jobs
WebSocket-style progress feed
No jobs yet.