YuE2-3B

YuE2-3B

Frontier music generation with editable scores

Developed by Multimodal Art Projection, YuE2 combines lyrics and musical style instructions to synthesize songs with singing and instrumental accompaniment.

• Score control: compose melody and harmony, melody alone, or generate without symbolic planning. An ABC score can be supplied directly.
• Revisions and covers: use an existing score with revised lyrics or style, or edit its melody and harmony before rendering again.
• Local generation: the upstream setup produces 48 kHz stereo audio on a 24 GB GPU without quantization.
• Pipeline access: separate interfaces expose planning and synthesis, with text guidance controls.

ComfyUI files

Packaged for ComfyUI by Comfy-Org, from the original M.A.P models.

• BF16 checkpoint: yue2_3b_bf16.safetensors (7.80 GB), the bfloat16 build.
• INT8 checkpoint: yue2_3b_int8_convrot.safetensors (3.96 GB), the smaller quantized build.
• SheetSage2 BF16 audio encoder: sheetsage2_bf16.safetensors (1.39 GB), an optional audio-to-score model for transcribing an existing recording before creating a cover. It is not required to generate a song from text and lyrics.

Local setup

Choose one checkpoint and place it in ComfyUI/models/checkpoints/. For audio-to-score workflows, place SheetSage2 in ComfyUI/models/audio_encoders/. See the ComfyUI package and SheetSage2 release for setup details.

Weights license: CC BY-NC 4.0.

Original release · Usage examples · Demos · GitHub · Discord