ACE-Step 1.5 XL launches with 4B music DiT

ACE-Step 1.5 XL represents a significant leap in open-source audio generation, doubling the parameter count of its predecessor to rival proprietary music models.

• –The 4B DiT decoder drastically improves audio richness and structural coherence, supporting full song generation in under 10 seconds on consumer-grade hardware.
• –Native support for lyrics-to-vocal, style-based LoRA training, and 50+ languages provides developers with a robust platform for custom AI music applications.
• –One-click RunPod templates and a pre-configured REST API server simplify the complex infrastructure required for high-VRAM model inference (12GB+ VRAM required).
• –Legal compliance with licensed and royalty-free training data makes the output viable for commercial use cases without copyright concerns.