Instructions to use arraypress/song-ace-step-1.5 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use arraypress/song-ace-step-1.5 with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("arraypress/song-ace-step-1.5", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - MLX
How to use arraypress/song-ace-step-1.5 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download arraypress/song-ace-step-1.5 --local-dir song-ace-step-1.5
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Model card
Browse files
README.md
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
base_model: ACE-Step/Ace-Step1.5
|
| 4 |
+
pipeline_tag: text-to-audio
|
| 5 |
+
tags:
|
| 6 |
+
- music-generation
|
| 7 |
+
- mlx
|
| 8 |
+
- ace-step
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# ACE-Step 1.5 for `song`
|
| 12 |
+
|
| 13 |
+
An unchanged copy of the files from
|
| 14 |
+
[ACE-Step/Ace-Step1.5](https://huggingface.co/ACE-Step/Ace-Step1.5) at revision
|
| 15 |
+
`19671f406d603126926c1b7e2adc169acbcade22` that the on-device `song` tool reads
|
| 16 |
+
(Swift + MLX, [swift-music-generation](https://github.com/arraypress/swift-music-generation)).
|
| 17 |
+
It is kept here so the tool keeps working if the upstream repository moves or changes.
|
| 18 |
+
Every file is byte-identical to upstream; `song` checks SHA-256 on download.
|
| 19 |
+
|
| 20 |
+
Only the 23 files the engine loads are included; the Python modelling code is not.
|
| 21 |
+
|
| 22 |
+
## What is inside
|
| 23 |
+
|
| 24 |
+
| Folder | Part | Upstream licence |
|
| 25 |
+
|---|---|---|
|
| 26 |
+
| `acestep-v15-turbo/` | DiT (turbo, 8 steps) and its silence latent | MIT (ACE-Step) |
|
| 27 |
+
| `acestep-5Hz-lm-1.7B/` | 5 Hz planning language model | MIT (ACE-Step) |
|
| 28 |
+
| `vae/` | Oobleck audio VAE | MIT (ACE-Step) |
|
| 29 |
+
| `Qwen3-Embedding-0.6B/` | text encoder | Apache-2.0 (Qwen) |
|
| 30 |
+
|
| 31 |
+
Licence texts are in `licenses/`. All credit for the model goes to the ACE-Step
|
| 32 |
+
team and, for the text encoder, the Qwen team.
|
| 33 |
+
|
| 34 |
+
## Citation
|
| 35 |
+
|
| 36 |
+
```bibtex
|
| 37 |
+
@misc{gong2026acestep,
|
| 38 |
+
title={ACE-Step 1.5: Pushing the Boundaries of Open-Source Music Generation},
|
| 39 |
+
author={Junmin Gong, Yulin Song, Wenxiao Zhao, Sen Wang, Shengyuan Xu, Jing Guo},
|
| 40 |
+
howpublished={\url{https://github.com/ace-step/ACE-Step-1.5}},
|
| 41 |
+
year={2026},
|
| 42 |
+
note={GitHub repository}
|
| 43 |
+
}
|
| 44 |
+
```
|