← 返回 Skills 市场

xeon-smartupscale_v2

Name: xeon-smartupscale_v2
Author: wray151

作者 Wray151 · GitHub ↗ · v0.1.7 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install xeon-smartupscale-v2

功能描述

Experiment-faithful 1080p upscaler for Intel Xeon CPUs. Decodes frames, resizes them in memory to 480x270 with lanczos, then runs Real-ESRGAN general-x4v3 x4...

使用说明 (SKILL.md)

Goal

Reproduce the exact 1080p experiment pipeline:

Decode source frames
In-memory lanczos -> 480x270
Real-ESRGAN general-x4v3 x4 -> 1920x1080
Encode with H.264 medium, crf 15

Usage

bash \x3Cskill_dir>/install.sh
bash \x3Cskill_dir>/smartupscale.sh \x3Cinput.mp4>

Optional custom output path:

bash \x3Cskill_dir>/smartupscale.sh \x3Cinput.mp4> -o \x3Coutput.mp4>

Packaging Notes

model/realesr-general-x4v3_480x270.xml and .bin are vendored for the fixed 1080p path.
No dynamic export path is included. This package stays on the fixed experiment model.
For ClawHub publishing, the raw .bin weights can be omitted from the package and restored from model/realesr-general-x4v3_480x270.bin.b64 or chunked model/realesr-general-x4v3_480x270.bin.b64.part-* files during install.sh.

Output

Final video: 1920x1080, H.264 medium, crf 15
Audio: copied from source when possible
Lanczos pre-scale is done in memory before inference; no encoded intermediate stage is produced.

安全使用建议

Install only if you are comfortable with the installer fetching and executing external code during setup. Prefer using a system-provided pip and ffmpeg/ffprobe already installed on the machine, or ask the publisher for pinned versions with checksum or signature verification before installing.

能力评估

✓ Purpose & Capability

The artifacts coherently implement a local 1080p video upscaling workflow using ffmpeg, OpenVINO, OpenCV, and bundled model weights; I found no unrelated credential access, broad file indexing, data upload, or persistence behavior.

ℹ Instruction Scope

Runtime instructions are narrow: install, then run smartupscale.sh on a user-supplied input video and optional output path. The docs disclose curl as a requirement, but do not clearly disclose that install.sh may execute get-pip.py from the network or download a static ffmpeg archive.

⚠ Install Mechanism

install.sh creates a venv and installs Python packages, but if pip is missing it executes curl -sSL https://bootstrap.pypa.io/get-pip.py | python; if ffmpeg is missing it downloads and extracts a prebuilt archive without checksum or signature verification. That is high-impact supply-chain execution under the user's privileges.

ℹ Credentials

The dependency footprint is plausible for video upscaling, but it is Linux x86_64-specific, downloads large/native tooling, and relies on external package sources during installation.

ℹ Persistence & Privilege

The skill persists local setup files under its own directory, including .venv, bin/ffmpeg, bin/ffprobe, and decoded model weights. I found no autorun hooks, background workers, cron/systemd entries, privilege escalation, or credential/session store access.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install xeon-smartupscale-v2
安装完成后，直接呼叫该 Skill 的名称或使用 /xeon-smartupscale-v2 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v0.1.7

Ship model weights as model/realesr-general-x4v3_480x270.weights.dat (base64) since ClawHub only accepts whitelisted text extensions

v0.1.6

Sidecar split to 300KB .weights.NNN.b64 chunks

v0.1.5

Chunk model sidecar into 350KB .weights.part-NN files to fit ClawHub size limit

v0.1.4

Rename model sidecar to .weights.b64 so ClawHub does not strip it

v0.1.3

Ship OpenVINO IR weights via base64 sidecar so ClawHub install delivers .bin

v0.1.2

Chunk model weights into .bin.b64.part files so ClawHub installs can reconstruct the Real-ESRGAN IR data

v0.1.1

Restore model weights from .bin.b64 so ClawHub installs include the Real-ESRGAN IR data

v0.1.0

Initial v2 release: exact 480x270 to 1920x1080 Real-ESRGAN experiment pipeline

元数据

Slug xeon-smartupscale-v2

版本 0.1.7

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 8

常见问题