spicystrippresident · Sep 28, 2026 · Sep 28, 2026 · Sep 28, 2026 · Sep 28, 2026 · Sep 28, 2026
diff --git a/LICENSING.md b/LICENSING.md
@@ -1,45 +1,65 @@
 # Licensing and source availability
 
-This document describes how Jasna's public source and optional supporter
-components are distributed. It does not replace the license notices in
-individual files or grant additional permissions.
+This document describes how Jasna's source, release packages, models, and
+optional supporter components are distributed. It does not replace the
+license notices in individual files or grant additional permissions.
 
-## Public source
+## Public application source
 
-Unless an individual file states otherwise, the source code in this repository
-is licensed under the GNU Affero General Public License version 3.0
-(`AGPL-3.0`). Files derived from other projects retain their copyright and
-license notices.
+Unless an individual file states otherwise, the source code in this
+repository is licensed under the GNU Affero General Public License version
+3.0 (`AGPL-3.0-only`). Files derived from other projects retain their
+copyright and license notices.
 
 Jasna includes code and model weights from
 [Lada](https://codeberg.org/ladaapp/lada). Those components remain covered by
 their applicable AGPL-3.0 notices. The vendored MMagic subset retains its
-OpenMMLab copyright and license notices.
+OpenMMLab copyright and Apache-2.0 notices. The LTX restoration code ported
+from Lightricks LTX-2.5 (`jasna/ltx/transformer.py`, `jasna/models/ltx_vae/`)
+and the LTX restoration model weights are covered by the LTX-2.x Community
+License.
 
-A clean checkout of this repository can be installed and used for Jasna's free
-restoration workflows without the optional `jasna.protection` package.
+The complete Jasna license is in [LICENSE](LICENSE). Copyright and attribution
+notices are in [NOTICE](NOTICE),
+[assets/THIRD_PARTY_LICENSES.md](assets/THIRD_PARTY_LICENSES.md), and
+[assets/THIRD_PARTY_MODELS.md](assets/THIRD_PARTY_MODELS.md).
+
+## Official release packages
+
+Official packages include the files above, the full license texts in
+`licenses/`, and the release-specific source correspondence in
+[RELEASE_SOURCES.md](RELEASE_SOURCES.md). The release page also provides a
+SHA-256 file for every split archive.
+
+The source tag and release package must have the same version. The tag is the
+preferred form for modifying Jasna's public application code. Exact custom
+dependency revisions, FFmpeg builds, patches, model hashes, and source links
+are recorded in `RELEASE_SOURCES.md` and the two third-party notice files.
 
 ## Optional supporter components
 
-Official release packages may additionally include a separately developed,
+Official release packages may include a separately maintained,
 source-unavailable protection component. It performs local supporter-key
-validation and in-memory decryption for the supporter-only `unet-4x` and
-SD 1.5 restoration models. When that component is unavailable, supporter-only
-models are unavailable, while the free restoration workflows remain usable.
+validation and in-memory decryption for supporter-only models. It contains no
+telemetry, persistence, or network communication.
 
-The protection component is maintained as a separate project, can be reused by
-other applications, and is distributed under separate proprietary terms. It
-is not included in the public Jasna source repository.
+The protection component is compiled into the official executable. There is
+no supported way to remove it from a prebuilt package. In the free CLI path,
+when no license arguments or supporter models are selected, Jasna does not
+import or call it. A public-source installation can run the free restoration
+workflows without the component.
 
-The protection component contains debugger and core-dump resistance intended
-to make extraction of model-decryption material more difficult. It does not
-implement telemetry, persistence, or network communication.
+The protection implementation is proprietary and its source is not included
+in the public repository. Consequently, the public source does not reproduce
+the official supporter-enabled executable bit for bit. The public repository
+must not be represented as complete corresponding source for that private
+component.
 
-## Release reproducibility
+## Models
 
-The public repository contains the application source and development setup
-for the free restoration workflows. It does not contain the proprietary
-protection component or the maintainer's standalone release tooling.
-Consequently, a public checkout does not reproduce a supporter-enabled
-official release package bit for bit.
+Every bundled or downloadable model has separate artifact-specific terms and
+a SHA-256 entry in `assets/THIRD_PARTY_MODELS.md`. Jasna-trained RF-DETR
+detectors are licensed under Apache-2.0. Lada models retain AGPL-3.0. Encrypted
+supporter models are proprietary and may be used only under their stated
+supporter-model terms.
 
diff --git a/NOTICE b/NOTICE
@@ -0,0 +1,19 @@
+Jasna
+Copyright 2026 Kruk2 and contributors
+
+Except where an individual file or bundled component states otherwise, Jasna
+is licensed under the GNU Affero General Public License version 3.0. The full
+license is in LICENSE.
+
+Jasna incorporates or distributes work from Lada, OpenMMLab MMagic,
+Lightricks LTX-2.5, Roboflow RF-DETR, PyAV, VALI, FFmpeg, VideoLAN, Python, and other projects.
+Their copyright, attribution, source, and license information is retained in:
+
+- LICENSING.md
+- RELEASE_SOURCES.md
+- assets/THIRD_PARTY_LICENSES.md
+- assets/THIRD_PARTY_MODELS.md
+- assets/licenses/
+
+Official binary packages also contain license files collected from the exact
+installed Python distributions under licenses/python-packages/.
diff --git a/README.ja.md b/README.ja.md
@@ -20,6 +20,7 @@ Jasna は無料です。支援者には、このプロジェクト用に訓練
 - [初回実行](#初回実行)
 - [さらに詳しく](#さらに詳しく)
 - [ベンチマーク](#ベンチマーク)
+- [ライセンス](#ライセンス)
 - [プロジェクトを支援する](#プロジェクトを支援する)
 - [謝辞](#謝辞)
 - [TODO](#todo)
@@ -45,7 +46,7 @@ Jasna は無料です。支援者には、このプロジェクト用に訓練
 
 - NVIDIA の **GTX 16 シリーズ / RTX 20 シリーズ以降**の GPU。GTX 10 シリーズ以前のカード（GTX 1050/1060/1070/1080）では動作しません。自分の GPU が対応か分からない場合は、NVIDIA の [GPU 一覧](https://developer.nvidia.com/cuda/gpus)で確認してください — コンピュート能力 7.5 以上が必要です。
 - Nvidia ドライバーは Windows で **610 以上**、Linux で **580 以上**。
-- AMD 対応は実験的で、ROCm 対応 GPU が必要です。
+- AMD 対応は実験的で、ROCm 対応 GPU（Radeon RX 7000/9000、Radeon 780M/760M/740M、Ryzen AI の内蔵 GPU）が必要です。Windows では AMD Adrenalin ドライバーを 26.8.1 以降に更新してください。
 - Jasna は、パスに英語の文字と数字のみを含むフォルダにインストールしてください。
 
 Jasna は VRAM を自動管理します。VRAM が不足すると、待機中のフレームを一時的にシステム RAM へ移します。設定は不要です。
@@ -143,6 +144,13 @@ RAM の中央値/ピーク、計測方法、生データ、パフォーマンス
 長時間動画の以前の結果は
 [従来のベンチマークレポート](benchmarks/2026-07-26_legacy_videos.md)（英語）にあります。
 
+## ライセンス
+
+Jasna の公開アプリケーションソースは AGPL-3.0 です。モデルと同梱依存関係には、
+それぞれのライセンスが適用されます。[ライセンスとソース公開範囲](LICENSING.md)、
+[モデルの通知](assets/THIRD_PARTY_MODELS.md)、
+[リリースソース対応表](RELEASE_SOURCES.md)をご覧ください。
+
 ## プロジェクトを支援する
 
 支援は追加モデルの訓練に使われます。主に、GPU のレンタル代と、より大きなデータセットで訓練するための計算時間です。支援者には以下を解除するキーが提供されます:
@@ -162,6 +170,11 @@ RAM の中央値/ピーク、計測方法、生データ、パフォーマンス
    - **[Unifans](https://app.unifans.io/c/kruk2)**: プラットフォームメッセージで送信されます。少し遅れる場合があります。
    - **[Buy Me a Coffee](https://buymeacoffee.com/kruk2)**、**暗号通貨**を含む: 支援時に使ったメールアドレスまたはハンドルへ送信されます。キーはそのメールアドレスまたはハンドルに紐付きます。
 
+キーは上記の2つのページで**のみ**販売しています。それ以外で Jasna、キー、改変ファイルを販売している人は詐欺です。そのキーは公式版では使えず、そのファイルはPC上で隠れたコードを実行する可能性があります。Jasna は GitHub からのみダウンロードしてください。
+
+有効化できない場合は、購入時のメールアドレスまたはユーザー名を確認し、キー全体をコピーし直してください。
+キーが無効というだけでは不正の証拠にはなりません。確認済みの非公式販売者のキーは別のメッセージで通知します。
+
 ## 謝辞
 
 - **[Lada](https://codeberg.org/ladaapp/lada)（Codeberg）** — Jasna は Lada に

diff --git a/README.md b/README.md
@@ -22,6 +22,7 @@ Jasna is free. Supporters get a key that unlocks the extra models trained for th
 - [First Run](#first-run)
 - [Learn More](#learn-more)
 - [Benchmarks](#benchmarks)
+- [Licensing](#licensing)
 - [Supporting the Project](#supporting-the-project)
 - [Acknowledgments](#acknowledgments)
 - [TODO](#todo)
@@ -47,7 +48,7 @@ Join the [SLS Discord](https://discord.gg/uNwQ4mHqgv) for examples, support, and
 
 - An NVIDIA **GTX 16-series / RTX 20-series or newer** GPU. GTX 10-series and older cards (GTX 1050/1060/1070/1080) won't work. Not sure about yours? Check NVIDIA's [GPU table](https://developer.nvidia.com/cuda/gpus) — compute capability 7.5+ is required.
 - Nvidia driver **610 or newer** on Windows, **580 or newer** on Linux.
-- AMD support is experimental and needs a ROCm-supported GPU.
+- AMD support is experimental and needs a ROCm-supported GPU (Radeon RX 7000/9000, Radeon 780M/760M/740M, Ryzen AI iGPUs). On Windows, update the AMD Adrenalin driver to 26.8.1 or newer.
 - Install Jasna into a folder whose path contains only English letters and numbers.
 
 Jasna manages VRAM automatically: when it runs low, waiting frames are temporarily moved to system RAM. No configuration needed.
@@ -146,6 +147,13 @@ the [full benchmark report](benchmarks/2026-07-26_release_matrix.md).
 Older full-length video results are in the
 [legacy benchmark report](benchmarks/2026-07-26_legacy_videos.md).
 
+## Licensing
+
+Jasna's public application source is AGPL-3.0. Models and bundled dependencies
+retain their own terms. See [licensing and source availability](LICENSING.md),
+[model notices](assets/THIRD_PARTY_MODELS.md), and the
+[release source correspondence](RELEASE_SOURCES.md).
+
 ## Supporting the Project
 
 Support pays for training extra models, mainly GPU rental and compute time for larger datasets. Supporters get a key that unlocks:
@@ -165,6 +173,11 @@ How to get a key:
    - **[Unifans](https://app.unifans.io/c/kruk2)**: sent by platform message. There might be a slight delay.
    - **[Buy Me a Coffee](https://buymeacoffee.com/kruk2)**, including **crypto**: sent to the email or handle used for the contribution. The key is tied to that email or handle.
 
+Keys are sold **only** on these two pages. Anyone else selling Jasna, keys or modified files is a scam: their keys do not work with the official app, and their files can run hidden code on your PC. Download Jasna only from GitHub.
+
+If activation fails, check the exact purchase email or handle and copy the full key again.
+An invalid key alone does not prove fraud; Jasna identifies keys from known unofficial sellers separately.
+
 ## Acknowledgments
 
 - **[Lada](https://codeberg.org/ladaapp/lada) (Codeberg)** — Jasna is inspired

diff --git a/README.zh.md b/README.zh.md
@@ -20,6 +20,7 @@ Jasna 是免费的。支持者会获得一个密钥，用于解锁为本项目
 - [首次运行](#首次运行)
 - [了解更多](#了解更多)
 - [基准测试](#基准测试)
+- [许可证](#许可证)
 - [支持本项目](#支持本项目)
 - [致谢](#致谢)
 - [TODO](#todo)
@@ -45,7 +46,7 @@ Jasna 是免费的。支持者会获得一个密钥，用于解锁为本项目
 
 - NVIDIA **GTX 16 系列 / RTX 20 系列或更新**的 GPU。GTX 10 系列及更旧的显卡（GTX 1050/1060/1070/1080）无法使用。不确定自己的显卡？请查看 NVIDIA 的 [GPU 表格](https://developer.nvidia.com/cuda/gpus) — 需要计算能力 7.5 或更高。
 - Nvidia 驱动: Windows 需 **610 或更新**，Linux 需 **580 或更新**。
-- AMD 支持是实验性的，需要 ROCm 支持的 GPU。
+- AMD 支持是实验性的，需要 ROCm 支持的 GPU（Radeon RX 7000/9000、Radeon 780M/760M/740M、Ryzen AI 核显）。Windows 上请把 AMD Adrenalin 驱动更新到 26.8.1 或更高版本。
 - 请把 Jasna 安装到路径只包含英文字母和数字的文件夹中。
 
 Jasna 会自动管理 VRAM: 显存不足时，等待中的帧会临时移动到系统内存。无需任何配置。
@@ -137,6 +138,13 @@ RAM 的中位数/峰值、测试方法、原始数据及性能提交对比，请
 较早的完整长视频结果见
 [旧版基准测试报告](benchmarks/2026-07-26_legacy_videos.md)（英文）。
 
+## 许可证
+
+Jasna 的公开应用程序源代码采用 AGPL-3.0。模型和捆绑依赖项保留各自的许可条款。
+请参阅[许可与源代码范围](LICENSING.md)、
+[模型声明](assets/THIRD_PARTY_MODELS.md)和
+[发行版源代码对应表](RELEASE_SOURCES.md)。
+
 ## 支持本项目
 
 支持用于训练额外模型，主要是租用 GPU，以及在更大数据集上训练所需的计算时间。支持者会获得一个密钥，用于解锁:
@@ -156,6 +164,11 @@ RAM 的中位数/峰值、测试方法、原始数据及性能提交对比，请
    - **[Unifans](https://app.unifans.io/c/kruk2)**: 通过平台消息发送，可能会有轻微延迟。
    - **[Buy Me a Coffee](https://buymeacoffee.com/kruk2)**，包括**加密货币**: 发送到贡献时使用的邮箱或账号。密钥与该邮箱或账号绑定。
 
+密钥**仅**在以上两个页面出售。其他任何出售 Jasna、密钥或修改版文件的人都是骗子：他们的密钥无法用于官方版本，他们的文件可能会在你的电脑上运行隐藏代码。请只从 GitHub 下载 Jasna。
+
+若激活失败，请核对购买时使用的邮箱或账号，并重新复制完整密钥。
+密钥无效本身并不能证明欺诈；Jasna 会单独提示来自已知非官方卖家的密钥。
+
 ## 致谢
 
 - **[Lada](https://codeberg.org/ladaapp/lada)（Codeberg）** — Jasna 受 Lada 启发，

diff --git a/RELEASE_SOURCES.md b/RELEASE_SOURCES.md
@@ -0,0 +1,79 @@
+# v0.11.0 release source correspondence
+
+This file records the custom and patched inputs used by official v0.11.0
+packages. The release tag is the source revision for Jasna itself. Release
+packages also include `LICENSE`, `LICENSING.md`, `NOTICE`, the third-party
+notices, and full license texts.
+
+## Source revisions
+
+| Component | Revision or source |
+| --- | --- |
+| Jasna | Git tag `v0.11.0`; the tag and package version must match |
+| Protection component | The private gitlink revision recorded by the `v0.11.0` tag; source is not published |
+| PyAV | Unmodified PyPI `av` 18.1.0, tag [`v18.1.0`](https://github.com/PyAV-Org/PyAV/commit/7e3d950a8b72062502c1a60d672f8ca565313af5) |
+| VALI | [`0e5c01ee57222a8ef4f9c7591284c4324bb3fba5`](https://codeberg.org/Kruk2/vali/commit/0e5c01ee57222a8ef4f9c7591284c4324bb3fba5) in the public Kruk2 fork |
+| RF-DETR | [`1.8.3` / `3bd6bffbcb13cac3a5b1c37da5a0fd5453b50c86`](https://github.com/roboflow/rf-detr/commit/3bd6bffbcb13cac3a5b1c37da5a0fd5453b50c86) |
+| MMagic compatibility patch | [`patches/fix_loading_mmengine_weights_on_torch26_and_higher.diff`](patches/fix_loading_mmengine_weights_on_torch26_and_higher.diff) |
+
+The VALI fork contains the complete modified source, including Jasna's decode
+status, corrupt-packet handling, timestamp, and decoder-construction changes.
+It is licensed under Apache-2.0.
+
+## FFmpeg in the PyAV and VALI wheels
+
+The PyAV wheels and the Linux VALI wheel link against the
+[`pyav-ffmpeg` 8.1.2-1](https://github.com/PyAV-Org/pyav-ffmpeg/releases/tag/8.1.2-1)
+build of FFmpeg 8.1.2. The source is FFmpeg tag `n8.1.2`, commit
+[`38b88335f99e76ed89ff3c93f877fdefce736c13`](https://github.com/FFmpeg/FFmpeg/commit/38b88335f99e76ed89ff3c93f877fdefce736c13).
+The exact build scripts are in the pyav-ffmpeg tag. They request x264/x265 and
+`--enable-version3`; because the wheel vendors and links the GPL-licensed x264
+library, Jasna distributes the combined wheel payload under GPLv3. FFmpeg's
+own runtime metadata identifies its libraries as LGPLv3-or-later.
+
+Linux x86-64 vendor archive:
+
+```text
+https://github.com/PyAV-Org/pyav-ffmpeg/releases/download/8.1.2-1/ffmpeg-manylinux-x86_64.tar.gz
+SHA-256 665b4903251dc753725e9cf64615c1af7708b932fb26bf16356acc77d76e95f8
+```
+
+## Bundled FFmpeg command-line tools
+
+Linux releases use the BtbN `linux64-gpl-8.1` build
+`ffmpeg-n8.1.2-34-g9b6c8969e0`. Its FFmpeg source is commit
+[`9b6c8969e05b4f0b29f0f85cd501be6b3e582e6b`](https://github.com/FFmpeg/FFmpeg/commit/9b6c8969e05b4f0b29f0f85cd501be6b3e582e6b).
+The BtbN GPL variant uses `--enable-gpl --enable-version3 --disable-debug`;
+the complete build scripts are in
+[BtbN/FFmpeg-Builds](https://github.com/BtbN/FFmpeg-Builds).
+
+```text
+https://github.com/BtbN/FFmpeg-Builds/releases/download/autobuild-2026-07-31-14-10/ffmpeg-n8.1.2-34-g9b6c8969e0-linux64-gpl-8.1.tar.xz
+SHA-256 09fc77be269c7053e438b7e96548e4af97604faf96a42c4a3c56a1ad74c22c0a
+```
+
+Windows packages use the GPLv3 FFmpeg 8 installation selected by the release
+builder. Before publishing, its `ffmpeg -version` output, source revision,
+configure line, and archive hash must be added to the release notes if they
+differ from the Linux build above.
+
+## Python and build environment
+
+Linux releases build CPython 3.13.14 from the official source archive:
+
+```text
+https://www.python.org/ftp/python/3.13.14/Python-3.13.14.tar.xz
+SHA-256 639e43243c620a308f968213df9e00f2f8f62332f7adbaa7a7eeb9783057c690
+```
+
+The application dependencies and exact direct version constraints are in
+`pyproject.toml`. Binary packages copy the license files from every installed
+Python distribution into `licenses/python-packages/`.
+
+## Reproducibility limit
+
+The public tag contains Jasna's public application source, public patches, and
+the source correspondence above. The final whole-application build driver and
+the protection implementation remain in the private protection project.
+Therefore a public checkout can reproduce the free workflows but cannot
+reproduce the supporter-enabled official executable bit for bit.