ACE Studio / StepFun · Music
The non-distilled fine-tune, on the same ace_step loader and conversion repo. It needs roughly three times the steps and higher guidance: its defaults are 27 steps at CFG 4.0, against the turbo's 8 at 1.0. In return it has the higher quality ceiling the turbo was distilled from. Try it when a turbo render is nearly right but thin.
Summarized from the linked source documents, last checked 25 August 2026. Every claim links to its source; anything unconfirmed is marked unconfirmed.
MIT. Commercial use permitted with no threshold. The project also states its training data was licensed, royalty-free, public-domain or synthetic. Training data is the usual licensing problem with music generators, and this statement is a claim on the model card, not a term of the licence.
MIT contains no output clause. The model card says generated music may be used commercially. Treat that as the publisher's stated intent. It is not a contractual grant, and there is no separate output-terms document.
Correction to the shipped catalog: The shipped catalog entry says "Apache-2.0 (unverified)". That is wrong on both counts: the LICENSE file in ace-step/ACE-Step-1.5 is MIT, and it is now verified. Both licences permit commercial use, so the practical answer was right, but the name was wrong. The shipped catalog also links to the older ace-step/ACE-Step repo; the 1.5 models live in ace-step/ACE-Step-1.5.
Based on the licence text as of the date below. Licences change; the links carry the authoritative version.
Transcribed from the shipped catalog, audiocpp_models_catalog.json — that file is authoritative if a number here disagrees with it.
audiocpp-ace-step-baseace_stepThere is no per-model install. One native binary runs all of these, so getting a model is a download of its weights and nothing else — no virtualenv, no torch pin, no CUDA wheel to match against your driver.
soundbox model install audiocpp
soundbox pull audiocpp-ace-step-base<models dir>/audiocpp/audiocpp-ace-step-base/ rather than in the shared Hugging Face cache, so a download stays scoped to the model you asked for.Primary sources: the model's own repository, the project it came from, and the conversion audio.cpp actually downloads. Never invented — an entry with nothing findable says so instead.
Licence details on these pages are summaries of the linked source documents, last checked 25 August 2026. They are not legal advice, and Soundbox is not affiliated with any model's authors — weights download from Hugging Face under the user's own account and terms. Company and model names identify whose work it is, and belong to their owners. The linked licence text is authoritative; commercial use should be checked against it, per model, before anything ships.