An AI-powered CLI tool for stable video object segmentation and removal using SVOR technology.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mi-visionforge-svor
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install mi-visionforge-svor using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is MiVisionForge-SVOR?
MiVisionForge-SVOR is a specialized skill designed for precise video object removal using Stable Video Object Removal (SVOR) technology. It allows developers and editors to identify, segment, and erase specific objects from video clips up to 3 seconds in length. By integrating with Openclaw Skills, this tool provides a streamlined command-line interface for complex visual tasks.
The skill utilizes a powerful combination of Vision-Language Models (VLM) for object suggestion and the MiEraser engine for high-quality inpainting. Whether you are dealing with common objects like pedestrians and cars or specific items that require coordinate-based targeting, this skill offers a robust solution for automated video cleanup and asset preparation.
MiVisionForge-SVOR Use Cases
- Removing background pedestrians or moving vehicles from cinematic shots.
- Erasing distracting signs, logos, or watermarks from video assets.
- Automating object isolation and mask generation for advanced visual effects.
- Cleaning up AI-generated video by removing unwanted artifacts or inconsistent elements.
- Preparing clean plates for motion tracking and compositing workflows.
How MiVisionForge-SVOR Works
- Initial segmentation identifies objects based on user-provided text prompts or labels.
- The system generates VLM recommendations, providing specific object IDs and bounding box coordinates.
- Users or agents select the target IDs to be removed based on the VLM analysis.
- The tool converts selected IDs into a binary mask where removal areas are set to pixel value 255.
- The MiEraser engine processes the video and mask to inpaint the target area, providing a final result via a secure URL.
MiVisionForge-SVOR Setup
To use this skill within the Openclaw Skills ecosystem, ensure your environment meets the following requirements:
# Check Python and FFmpeg dependencies
python --version
ffmpeg -version
# Install the requests library
pip install requests
# Configure your API Key
export SVOR_API_KEY="your_svor_api_key"
MiVisionForge-SVOR Data Schema & Taxonomy
The skill manages video data and metadata through specific CLI arguments and structured outputs:
| Argument | Description | Format |
|---|---|---|
| --video | Path to the input mp4 file | String |
| --classes | Definition of object labels and prompts | label:text:prompt |
| --targets | Specific object IDs for removal | label:1,2,3 |
| --boxes | Normalized coordinates for manual targeting | x,y,w,h (0-1) |
| --points | Precise point coordinates for targeting | x,y (0-1) |
| --frame | The specific frame index used for annotation | Integer (0-based) |
| --tmp-dir | Directory for temporary FFmpeg processing | Path |
MiVisionForge-SVOR Advanced Features
- VLM-assisted object analysis to distinguish between foreground and background elements.
- Multi-class object removal support using advanced mask blending techniques.
- Frame-specific targeting to handle objects that appear or become clear at specific timestamps.
- Support for both text-based prompts and coordinate-based (boxes/points) targeting for maximum precision.
- Automatic video compression and path handling to bypass gateway size limits and FFmpeg path restrictions.
SKILL.md
Loading
Related Openclaw Skills
Household Book OCR
A specialized OCR skill for Openclaw Skills that extracts structured data from household register documents and images.
Design Pattern Selector
An intelligent diagnostic tool for AI agents to identify and apply the optimal Gang of Four design pattern for complex software architecture challenges.
Skill Graphify
Skill Graphify transforms any folder of code, documents, or images into a navigable and queryable knowledge graph with automated community detection.
sbti-haojity: SBTI Cyber Personality Test
A technical implementation of the viral SBTI personality test allowing users and AI agents to discover their cyber-personality traits through a structured 30-question workflow.
MiniMax Media Generation
A unified MiniMax API wrapper for generating high-fidelity audio, cinematic images, and professional video clips through a single CLI interface.
OO Design Principle Evaluator
A specialized diagnostic tool that evaluates object-oriented architectures against foundational Gang of Four design principles to eliminate brittle inheritance and rigid coupling.
Featured*
