Token导航 LogoToken导航TokenDH.com
效率需要联网clawhub未标认证来源可访问clear审计通过

ai-video-stabilizerAI 视频稳定器

Agent Skill

用于辅助视频生成、动画合成、脚本化剪辑或 Remotion 等视频项目开发。它适合让 Agent 组织镜头、生成素材说明、维护合成代码或排查渲染问题。使用时需要确认分辨率、时长、素材路径和导出格式;涉及外部素材、人物肖像或商业发布时,应先核对版权授权和内容审核要求。

总安装

3,552

周安装

148

GitHub Stars

公开资料未说明

下载量

1,184
OpenClaw

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

MIT-0

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:ai-video-stabilizer(AI 视频稳定器)
来源仓库:https://github.com/peand-rover/ai-video-stabilizer
安装命令:
openclaw skills install ai-video-stabilizer
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install ai-video-stabilizer

简介

用于修复手持抖动与相机运动异常,实现视频平滑稳定处理。

  • 支持行走弹跳、风摇等多种抖动类型修复,提升画面观感质量。
  • 使用时需确认输入视频格式与版权,涉及商业用途应审核合规性。
  • 安装方式:通过 clawhub 安装,命令为 openclaw skills install ai-video-stabilizer。
  • 注意:涉及人物或品牌内容时,需额外核对授权与真实性。

SKILL.md

name
ai-video-stabilizer
version
1.0.1
displayName
AI Video Stabilizer — Stabilize Shaky Video and Smooth Camera Motion with AI
description
>
metadata
{"openclaw": {"emoji": "📹", "requires": {"env": [], "configPaths": ["~/.config/nemovideo/"]}, "primaryEnv": "NEMO_TOKEN"}}
homepage
https://nemovideo.com
repository
https://github.com/nemovideo/nemovideo_skills

AI Video Stabilizer — Smooth Footage from Any Camera, Any Situation

Shaky footage is the most common quality problem in video. Gimbal stabilizers cost $100-500 and add weight, setup time, and another device to charge. Tripods eliminate shake but eliminate mobility. Most real-world recording happens handheld: phone in hand while walking, camera at a concert, action cam on a bike, drone in wind, car dashcam on rough roads. All of these produce shake that makes footage look amateur and can cause viewer discomfort. The human eye has built-in stabilization — the vestibulo-ocular reflex keeps our visual field steady even when our head moves. When we watch shaky footage, our brain tries to stabilize it and fails, creating a disconnect that registers as low quality (and for some viewers, actual motion sickness). Professional stabilization in post-production traditionally requires: importing into editing software, applying warp stabilizer or similar tool, waiting for frame-by-frame analysis (minutes to hours depending on length), adjusting settings to balance smoothness vs. crop, re-rendering, and reviewing. For long footage, the process can take hours. NemoVideo stabilizes through AI analysis that distinguishes intentional movement (pans, tilts, tracking shots) from unintentional shake (hand tremor, footstep bounce, vehicle vibration). The AI removes the shake while preserving the creative movement — producing footage that looks like it was shot on a gimbal.

Use Cases

  1. Walking Footage — Eliminate Footstep Bounce (any length) — A vlogger records while walking through a city. Every step creates a vertical bounce that makes the footage feel like a shaky-cam horror movie. NemoVideo: analyzes the repetitive vertical oscillation pattern (footstep frequency), removes the bounce while preserving the forward walking motion, maintains horizontal panning (the vlogger looking at different things), and produces footage that feels like a smooth dolly shot. Walking footage that looks like a steadicam.
  1. Concert/Event — Handheld in a Crowd (any length) — Phone footage from a concert: the shooter is being jostled by the crowd, holding their phone above their head, zoomed in on the stage. Triple shake: crowd pushing, arm fatigue, and digital zoom amplifying every movement. NemoVideo: applies aggressive stabilization appropriate for the extreme shake level, crops slightly to allow stabilization headroom (the AI determines the minimum crop needed), preserves the audio perfectly (stabilization is video-only), and produces watchable footage from nearly unwatchable source material.
  1. Car/Vehicle — Dashcam Vibration (any length) — Dashboard camera footage from a road with potholes and rough pavement. High-frequency vibration makes every frame slightly blurry and the video uncomfortable to watch. NemoVideo: identifies the high-frequency vibration pattern (different from hand shake — faster, more regular), removes the vibration while keeping the vehicle's actual turning and direction changes, and produces smooth driving footage. Dashcam footage that actually documents the drive clearly.
  1. Drone — Wind Wobble Correction (any length) — Drone footage in moderate wind. The drone's gimbal handles most stabilization but wind gusts create periodic wobbles — slight tilts and position shifts that the gimbal cannot fully compensate. NemoVideo: detects the irregular wobble pattern (wind gusts are aperiodic, unlike footstep bounce), smooths the wobbles while preserving intentional flight path and camera movements, and produces footage that looks like calm-weather flight. Drone footage usable regardless of wind conditions.
  1. Action Cam — Extreme Motion Stabilization (any length) — GoPro mounted on a mountain bike, ski helmet, or surfboard. Extreme vibration and rotation that built-in stabilization cannot fully handle. NemoVideo: applies maximum stabilization with AI-powered motion prediction (predicting where the frame should be based on surrounding frames), accepts higher crop ratio for extreme cases (the crop needed for extreme stabilization is compensated by the action cam's wide-angle lens), and produces smooth action footage suitable for highlight reels and social sharing.

How It Works

Step 1 — Upload Shaky Video

Any footage with unwanted camera shake. NemoVideo auto-detects the type and severity of shake.

Step 2 — Set Stabilization Level

Auto (AI decides), gentle (minimal correction, minimal crop), standard (balanced), aggressive (maximum smoothness, more crop), or custom (specify smoothing strength and crop tolerance).

Step 3 — Generate

curl -X POST https://mega-api-prod.nemovideo.ai/api/v1/generate \
  -H "Authorization: Bearer $NEMO_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "skill": "ai-video-stabilizer",
    "prompt": "Stabilize a 5-minute handheld walking vlog. Remove footstep bounce and hand shake but preserve intentional panning when I look at things. Standard stabilization — balance between smoothness and crop. Maintain original resolution after crop (scale up slightly). Also generate a 9:16 vertical version with face tracking for TikTok clips.",
    "stabilization": "standard",
    "preserve_motion": ["pan", "tilt", "zoom"],
    "remove_motion": ["shake", "bounce", "vibration"],
    "crop_handling": "scale-to-fill",
    "outputs": [
      {"format": "16:9", "resolution": "1080p"},
      {"format": "9:16", "tracking": "face", "platform": "tiktok"}
    ]
  }'

Step 4 — Compare Before/After

Preview stabilized vs. original side-by-side. Verify: shake is removed, intentional movement is preserved, no warping artifacts. Download.

Parameters

ParameterTypeRequiredDescription
promptstringStabilization requirements
stabilizationstring"auto", "gentle", "standard", "aggressive", "custom"
smoothingfloat0.0-1.0 custom smoothing strength
preserve_motionarray["pan", "tilt", "zoom", "tracking"] — intentional motions to keep
remove_motionarray["shake", "bounce", "vibration", "roll"]
crop_handlingstring"scale-to-fill", "black-bars", "ai-extend"
max_cropfloatMaximum crop percentage (default: 15%%)
outputsarray[{format, resolution, tracking}]
batcharrayMultiple videos

Output Example

{
  "job_id": "avstab-20260328-001",
  "status": "completed",
  "source_duration": "5:12",
  "shake_analysis": {
    "type": "walking-bounce + hand-shake",
    "severity": "moderate",
    "frequency": "2.1 Hz (footstep) + random (hand)"
  },
  "stabilization_applied": "standard",
  "crop_used": "8.2%%",
  "intentional_motion_preserved": ["horizontal pan", "vertical tilt"],
  "outputs": {
    "main": {"file": "vlog-stabilized-16x9.mp4", "resolution": "1920x1080"},
    "tiktok": {"file": "vlog-stabilized-9x16.mp4", "resolution": "1080x1920", "tracking": "face"}
  }
}

Tips

  1. Standard stabilization is right for 90% of footage — Gentle leaves visible shake. Aggressive crops too much. Standard balances smoothness and frame preservation for most handheld recordings.
  2. Preserving intentional motion is what separates good stabilization from bad — A stabilizer that removes everything (pans included) produces eerie, floating footage. NemoVideo's AI distinguishes shake (high-frequency, random) from pans (smooth, directional) and removes only the shake.
  3. Wider lens = easier stabilization — Wide-angle footage (phone, GoPro, drone) has more frame to work with for cropping. Zoomed-in telephoto footage amplifies shake and leaves less crop margin. When possible, shoot wide and crop in post.
  4. Scale-to-fill after crop maintains resolution — Stabilization requires slight cropping. Scaling the stabilized result back to full resolution prevents visible borders while maintaining apparent resolution. The slight softening from upscaling is invisible at normal viewing distances.
  5. Audio is never affected by stabilization — Video stabilization only modifies the visual frames. Audio remains perfectly synced and unaltered. No need to worry about audio drift or quality changes.

Output Formats

FormatResolutionUse Case
MP4 16:91080p / 4KYouTube / website
MP4 9:161080x1920TikTok / Reels / Shorts
MP4 1:11080x1080Instagram / LinkedIn

Related Skills

适合场景

01

OpenClaw 用户查找和安装 Skill 时

02

用户想查找某类 Agent Skill 时

03

需要根据任务场景推荐可安装能力包时

04

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

补充不同宿主或平台的使用分布数据

能力 5

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

OpenClaw

80.4%
按下载量换算952

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills