YuE2-3B BF16 baked with YuE2-instrumental-cot-full-lora
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.
Files
Packaged for YuE2 UI by LeeAeron and ComfyUI by Comfy-Org, from the original M.A.P models.
BF16 checkpoint: 6.67 Gb, BF16 baked model, packaged into GGUF.
Additional info:
1. Bare — let the model choose the structure and length.
[instrumental]
2. Untimed tags — you choose the section order, the model chooses the timing.
[intro]
[verse]
[chorus]
[bridge]
[chorus]
[outro]
3. Timed tags — you also give each section a start and end in m:ss. The model was trained with exact section times from real tracks, so this is the
strongest structural steer. Treat the times as a guide rather than a guarantee: the model follows the section order and proportions better than the
absolute end time, and it tends toward 3–5 minute songs regardless of the plan.
[intro 0:00-0:15]
[verse 0:15-0:45]
[chorus 0:45-1:10]
[bridge 1:10-1:40]
[chorus 1:40-2:05]
[outro 2:05-2:30]
Keep production notes [guitar solo, distorted]), lyrics text, and literal \n characters out of the lyrics field: baked LoRA never saw them and they pull the output away from the trained behavior.
Also, availabe in HugginFace:
Weights license: CC BY-NC 4.0.