Clarify weight and documentation license scopes

LICENSE

@@ -1,3 +1,18 @@
+ACappellaSet licensing scope
+
+Model checkpoints: Creative Commons Attribution-NonCommercial 4.0
+International (CC BY-NC 4.0). Full weight license:
+licenses/CC-BY-NC-4.0.txt
+
+Upstream notices and third-party rights remain applicable; see
+WEIGHTS_NOTICE.md.
+
+The MIT License below applies only to original repository documentation and
+metadata. It does not license model checkpoints. Source code is maintained
+separately on GitHub under MIT with third-party notices.
+
+-----------------------------------------------------------------------
+
 MIT License
 
 Copyright (c) 2026 ACappellaSet contributors

README.md

@@ -11,10 +11,11 @@ inference: false
 
 # ACappellaSet models
 
+**Licensing scope:** All 29 model checkpoints use **[CC BY-NC 4.0 — full weight license](licenses/CC-BY-NC-4.0.txt)**, subject to [upstream notices and third-party rights](WEIGHTS_NOTICE.md). The root [LICENSE](LICENSE) applies MIT only to original documentation and metadata; it does not license the model weights. [Source code on GitHub](https://github.com/pting126/ACappellaSet) uses MIT with third-party notices.
+
 Model weights and release metadata for a cappella source separation.
 
 The public release contains **29 checkpoints**: three fixed-output models, twelve original part-specific specialists, and fourteen conditioned selections, including two trained with AI-generated singer augmentation. Downloads do not require login or an access request.
-**Weight license: [CC BY-NC 4.0](licenses/CC-BY-NC-4.0.txt)**, subject to the upstream notices and third-party rights described in [WEIGHTS_NOTICE.md](WEIGHTS_NOTICE.md).
 
 - **Checkpoint paths, hashes, configurations, and validation selections:** [manifest.json](manifest.json)
 - **Code and installation:** [ACappellaSet on GitHub](https://github.com/pting126/ACappellaSet)
@@ -91,7 +92,7 @@ Validation-selected models record the validation corpus, shared/inclusive scope,
 
 ## Licenses and citation
 
-Original repository documentation and metadata use the [MIT license](LICENSE).
+Original repository documentation and metadata use the [MIT license](LICENSE), which does not apply to model weights.
 All 29 model checkpoints use [CC BY-NC 4.0](licenses/CC-BY-NC-4.0.txt). Preserve attribution and indicate changes when sharing. This project license applies only to rights the licensors can grant; upstream notices and third-party rights remain applicable. See [weight notices](WEIGHTS_NOTICE.md). Dataset audio is separate.
 Upstream architecture attribution is maintained in the GitHub repository's
 [third-party notices](https://github.com/pting126/ACappellaSet/blob/main/THIRD_PARTY_NOTICES.md).

WEIGHTS_NOTICE.md

@@ -14,5 +14,5 @@ B6-LM encoder weights; preserve the upstream [Apache license](licenses/WeSpeaker
 and [WeSpeaker attribution](https://huggingface.co/Wespeaker/wespeaker-voxceleb-redimnet2-B6-LM).
 Project fine-tuning changes pretrained tensors. All released checkpoints are inference-only exports, retaining every separator and matching encoder tensor while omitting optimizer state, training arguments, and internal filesystem paths.
 
-Original documentation and metadata retain their [MIT license](LICENSE).
+Original documentation and metadata retain their [MIT license](LICENSE); that MIT license does not apply to model weights.
 Dataset audio, corpus terms, and third-party rights remain separate.