Who makes ACE-Step 1.5 XL?

ACE-Step 1.5 XL is made by ACE Studio.

Does ACE-Step 1.5 XL work in the EU?

Available in the EU. Weights and code are MIT-licensed and downloadable from GitHub and Hugging Face with no regional restriction, so they can be run locally in the EU. No explicit EU data-residency statement was found for the hosted acemusic.ai service. Last verified 14 Nov 2026.

How much does ACE-Step 1.5 XL cost?

Open source, open source. Free to download and run locally; a free hosted version is offered at acemusic.ai per the README. Licence: MIT.

Where was ACE-Step 1.5 XL shown?

In 1 creator video by AI Search. Most recently by AI Search on 15 Aug 2026, at 17:49 in "The BEST local AI music generator is here!".

ACE-Step 1.5 XL

by ACE Studio

Try it Website Docs GitHub Hugging Face

Open-source local music generation model with a 4B-parameter DiT decoder, extending ACE-Step 1.5 with higher audio quality.

Available in EUMusicModel

Updated 14 Nov 2026EU status checked 14 Nov 2026Last shown 15 Aug 2026

ACE-Step 1.5 XL at a glance

What is ACE-Step 1.5 XL?: ACE-Step 1.5 XL is a model from ACE Studio. Open-source local music generation model with a 4B-parameter DiT decoder, extending ACE-Step 1.5 with higher audio quality.
Who makes ACE-Step 1.5 XL?: ACE-Step 1.5 XL is made by ACE Studio.
Does ACE-Step 1.5 XL work in the EU?: Available in the EU. Weights and code are MIT-licensed and downloadable from GitHub and Hugging Face with no regional restriction, so they can be run locally in the EU. No explicit EU data-residency statement was found for the hosted acemusic.ai service. Last verified 14 Nov 2026.
How much does ACE-Step 1.5 XL cost?: Open source, open source. Free to download and run locally; a free hosted version is offered at acemusic.ai per the README. Licence: MIT.
Where was ACE-Step 1.5 XL shown?: In 1 creator video by AI Search. Most recently by AI Search on 15 Aug 2026, at 17:49 in “The BEST local AI music generator is here!”.

Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.

About

ACE-Step 1.5 XL is the larger variant of ACE-Step 1.5, an open-source music foundation model. It uses a hybrid design in which a language model plans the song (metadata, lyrics, captions via chain-of-thought) and a Diffusion Transformer renders the audio. The XL series, announced on 2026-04-02, swaps in a 4B-parameter DiT decoder for higher audio quality and is released in three variants: xl-base, xl-sft and xl-turbo. The existing LM models are stated to be fully compatible.

The model generates full songs from text descriptions or lyrics, with durations from 10 seconds to 10 minutes, batch generation of up to 8 songs, and lyrics in 50+ languages. Editing and control features include reference-audio guidance, cover generation, repainting, track separation, multi-track layering, vocal-to-accompaniment, and control of BPM, key and time signature. LoRA training for personal styles is supported through the Gradio UI.

It runs locally through a Gradio UI or REST API server (Python 3.11-3.12, CUDA recommended; also MPS, ROCm, Intel XPU, CPU). The XL models need at least 12 GB VRAM with offloading and 20 GB or more is recommended. The project also points to a free hosted service at acemusic.ai. The original ACE-Step project page lists limitations such as output inconsistency across seeds, continuity artifacts in repaint/extend, and coarse vocal nuance; these were written for v1 and may not fully apply to 1.5.

Key features