Repository navigation
Releases: PengJunchen/dsh-video-studio
Release list
DSH Video Studio v0.1.0-beta.3 · 自带环境的跨平台安装包
DSH Video Studio · 影调工作台
Complete per-platform installation packages now include Python 3.12.16, FFmpeg/ffprobe 9.0.2, NumPy 2.2.6, OpenCV 4.11 and all 18 DJI restoration LUTs. Select the suffix matching the Harness Host and its Node architecture:
| Download suffix | System |
|---|---|
| darwin-arm64 | macOS 13+, Apple Silicon |
| darwin-x64 | macOS 13+, Intel |
| win32-x64 | Windows 10 22H2+, x64 |
| linux-x64 | Ubuntu 22.04+ / glibc 2.35+, x86-64 |
The package without a platform suffix is the small universal package: it includes LUTs and provides an explicit first-run environment setup button. Complete platform packages need no separate Python, FFmpeg, pip, Homebrew, compiler, or startup runtime download. DeepSeek Harness provides the plugin host and Node.
- Inline environment detection and version display; independent setup script for the universal package.
- Fixed runtime archives and original LUT bytes are checked against pinned hashes and sizes.
- Per-process Python/PATH isolation, portable Windows LUT paths and UTF-8 handling.
- Each complete package is extracted into a fresh directory and checked for scan, LUT/tone/denoise preview, 10-bit HEVC export, and private Python/CRT use on its native OS before release.
- Original LUTs come from DJI official downloads; sources, versions and hashes are included.
完整系统包已内置固定 Python、FFmpeg/ffprobe、人物处理依赖和实际 LUT。下载对应系统后缀的 .tgz，在 Harness 插件安装界面选择该文件即可。轻量通用包提供首次启动的一键独立环境配置。完整包启动无需额外下载环境。
Mac 版已在本机完成隔离目录、真实 D-Log2 Vivid 65 预览及进程禁网条件下的扫描、预览与导出验证；四个平台的发行检查由对应系统的 CI 执行。
DSH Video Studio v0.1.0-beta.2 · 完整 LUT 安装包
DSH Video Studio · 影调工作台
Download dsh-video-studio-0.1.0-beta.2.tgz below for installation in DeepSeek Harness. GitHub-generated source ZIP/tar archives require resource preparation first.
- Includes 18 original DJI restoration LUTs, 44 exact camera/profile/look mappings and compatible size variants. LUTs come from DJI official downloads; original URLs, versions and hashes are recorded in
assets/manifest.json. - Automatically discovers the packaged resources. Prefers Vivid / Size 65 when available for the exact camera/profile, with user-selectable alternatives.
- Source builds check archive and per-file SHA-256, size and grid before packing. Online, original-DJI and offline build routes are documented in the bilingual READMEs and
assets/README.md. - Python 3.10+ and FFmpeg/ffprobe 8+ remain Host requirements; optional NumPy/OpenCV enables face-guided skin processing. No video files or neural model weights are bundled.
下载下方完整 .tgz 后，在 Harness 插件安装界面选择该文件。安装包包含 DJI 官方来源的实际 LUT，安装后无需手工配置资源路径。默认优先对应机型与格式的 Vivid／Size 65，仍可切换。Python 和 FFmpeg 由宿主环境提供。
构建步骤、资源包打包脚本及依赖准备均在仓库提供；源码采用 MIT 协议。
v0.1.0-beta.1
First public beta of DSH Video Studio, a local video color-editing plugin for DeepSeek Harness.
- One installable package with a shared runtime, the Studio and five feature entries.
- Camera LUT matching, independent highlights/shadows, optional low-light lift and spatial denoising, face-guided skin treatment, selective color and detail-first export.
- English and Chinese README, MIT-licensed original source. Third-party LUTs, footage, model weights and FFmpeg binaries are not bundled.
Install the attached .tgz in Harness, or use github:PengJunchen/dsh-video-studio#v0.1.0-beta.1. Requires Node 22.19+, Python 3.10+ and FFmpeg/ffprobe 8+. Face-guided treatment additionally needs NumPy/OpenCV.
Early-release limits: SDR Rec.709 workflow; original-size 4K processing is not real-time; tone suggestions use sparse samples; denoising is spatial and may soften fine texture.
影调工作台首个公开测试版
一个安装包提供统一工作台、五个功能入口和共享处理服务。包含相机 LUT 匹配、独立高光／暗部调整、可选暗光增强与空间降噪、人物肤色、局部颜色和细节优先导出。
附中英文 README，原创源码采用 MIT 协议。第三方 LUT、原视频、模型权重和 FFmpeg 二进制不随包提供。下载 .tgz 后可在 Harness 插件界面安装；运行要求及配置方法见 README。
当前为早期版本：使用 SDR Rec.709 工作域，原尺寸 4K 处理未达到实时；建议基于少量抽样，空间降噪可能柔化细弱纹理。
Video Studio fixed runtimes · media-runtime-20261029-1
Fixed per-platform Python, FFmpeg and vision dependencies. See runtime/README.md and the included input manifests for build sources and platform requirements.
DJI restoration LUT resources v2
DJI restoration LUTs from DJI official downloads. Original pages, URLs, versions, sizes and SHA-256 hashes are recorded in the project resource manifests.
Contains 18 original 3D CUBE files, 44 camera/profile/look mappings and compatible size variants. The archive is built with scripts/package-resources.py; original LUT bytes are preserved.
LUT 来自 DJI 官方下载。原始页面、地址、版本、尺寸和散列见项目资源清单。含 18 个原始 3D CUBE 文件、44 条映射及尺寸变体；使用仓库中的资源打包脚本构建，不改写 LUT。
DJI restoration LUT resources v1
Locked DJI restoration resources for DSH Video Studio.
18 original 3D CUBE payloads, 44 exact camera/profile/look mappings, compatible size variants, and source/distribution notices. Each payload is content-addressed by SHA-256. The application source manifest pins this archive by SHA-256; its original official DJI URLs remain recorded in the application source.
The maintainer has confirmed authorization for inclusion in this project distribution. These are DJI materials, not MIT-licensed project code; see the included NOTICE.md.
影调工作台固定资源包：含 18 个 DJI 原始还原 LUT、44 条精确映射和尺寸变体。资源用于完整安装包及可复现构建，不包含视频或模型权重。维护者已确认可随本项目分发，DJI 原始资源不属于项目 MIT 授权。