Community (BS-RoFormer) · Utility

BS-RoFormer stem separation

Terms unconfirmedUnconfirmed (MIT code; checkpoint licence undeclared)

The weakest licence provenance in this catalog. The reference implementation and the widely used training repo are both MIT, but the checkpoints are listed by author across GitHub releases and Hugging Face with no licence stated for any of them. Third-party mirrors fill the gap with tags of their own. This page reports the licence as unknown.

The licence

Summarized from the linked source documents, last checked 25 August 2026. Every claim links to its source; anything unconfirmed is marked unconfirmed.

Unconfirmed (MIT code; checkpoint licence undeclared)

Terms unconfirmed

No licence could be confirmed for these weights. The reference implementation and the training repo are both MIT, and audio.cpp's conversion table labels this package Apache-2.0. The upstream checkpoint list declares no licence for any of its models, and no statement from whoever trained this checkpoint could be found. The two secondary labels also disagree with each other.

Conditions

  • Treat as unlicensed until you get an answer from the checkpoint's author. Some mirror cards argue that a release asset inherits its repository's MIT licence; no upstream source says so.

What about the audio you generate?

No output clause found anywhere in this family. As with any separator, what you may do with the stems is governed mainly by your rights in the audio you fed in.

Correction to the shipped catalog: The shipped catalog says "Unknown", and that is still accurate. The reason is now specific: the people distributing these checkpoints have never declared a licence. audio.cpp's table calls this package Apache-2.0, which conflicts with the MIT reference code and the undeclared checkpoint. Those could not be reconciled, so no claim is made.

Specs

Transcribed from the shipped catalog, audiocpp_models_catalog.json — that file is authoritative if a number here disagrees with it.

Vendor
Community (BS-RoFormer)
Catalog id
audiocpp-bs-roformer-sep
audio.cpp loader family
bs_roformer
Can do
Separation
Weights file
BS-RoFormer-ep368-GGUF/bs-roformer-ep368-q8_0.gguf
Account needed to download
No
Max clip length
600s
Sample rate
44.1kHz stereo
Approx. VRAM
3.0 GB
In the picker by default
No — opt in from Settings
Run end-to-end here
No — catalogued, not tested here

Getting it

There 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.

From the CLI
soundbox model install audiocpp
soundbox pull audiocpp-bs-roformer-sep
  • The binary is fetched from audio.cpp's GitHub releases the first time you ask for it and cached under your own data directory. Soundbox ships no native binaries and vendors none.
  • Weights land in <models dir>/audiocpp/audiocpp-bs-roformer-sep/ rather than in the shared Hugging Face cache, so a download stays scoped to the model you asked for.
  • This one is opt-in: switch it on in the Models page's “Enable more models” section before it appears in the picker.

Worth knowing

  • Untested end-to-end here. Both the weights glob and the licence need confirming before real use.
  • Same input rules as htdemucs: source goes in as the audio input, and Soundbox resamples to the model's rate first.

Sources

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.