Updated: May 16, 2025
base modelDownload
1 variant available
fp16 SafeTensor
ace_step_v1_3.5b.safetensors
Half precision, best balance β’ 7.17 GB
Verified: a year ago
1910 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 90 1 2 3 4 5 6 7 8 9
(14)
May 7, 2025
Other

1K0 1 2 3 4 5 6 7 8 9K
2.9K0 1 2 3 4 5 6 7 8 9.0 1 2 3 4 5 6 7 8 9K

License:
ACE-Step
A New Era in AI-Driven Music Generation
workflow https://civitai.com/models/1557004
π Highlights
Blazing Speed
Up to 15Γ faster than LLM-based baselines
~20 s for a 4 min track on NVIDIA A100
Holistic Architecture
Diffusion-based Synthesis
Deep Compression AutoEncoder (DCAE)
Linear Transformer for long-range control
Full-Song Generation
Durations from 30 s to 5 min
Natural-language prompts + lyric tags (
[verse],[chorus])
πΆ Strengths
Musical Coherence
Balanced melody, harmony & rhythmControllability
Duration, style, structureStyle Transfer
Remix genres or emulate artist textures
π€ Ethical Use
βοΈ Verify originality
βοΈ Disclose AI assistance
βοΈ Respect cultural & copyright norms
π« Avoid harmful or deceptive content
License: Apache 2.0
Developed by: ACE Studio & StepFun
Official Repo: ACE-Step v1-3.5B on Hugging Face
