import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("AEmotionStudio/acestep-models", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]ACE-Step 1.5 Models (FFMPEGA Mirror)
Mirror of ACE-Step/Ace-Step1.5 models for use with ComfyUI-FFMPEGA.
Contents
| Directory | Size | Description |
|---|---|---|
acestep-v15-turbo/ |
~4.8 GB | DiT turbo model (8-step inference) |
acestep-5Hz-lm-1.7B/ |
~3.8 GB | 1.7B language model for lyric planning |
acestep-5Hz-lm-0.6B/ |
~1.4 GB | 0.6B language model (lower VRAM option) |
Qwen3-Embedding-0.6B/ |
~1.2 GB | Text encoder |
vae/ |
~0.3 GB | VAE decoder |
Usage
These models are automatically downloaded by ComfyUI-FFMPEGA's ACE-Step integration. Place the checkpoints/ contents in:
ComfyUI/models/acestep/checkpoints/
License
MIT — same as the original ACE-Step 1.5 release.
Credits
- Original model: ACE-Step by the ACE-Step team
- Mirror maintained by AEmotionStudio
- Downloads last month
- -