README.md

ModelScope with VideoEraser

VideoEraser integration with ModelScope/ZeroScope for concept erasure in text-to-video generation.

Quick Start

cd ModelScope

python inference.py \
    --model cerspense/zeroscope_v2_576w \
    --prompt "A man running under starry night by Van Gogh." \
    --erased-concept "Van Gogh" \
    --output ./outputs \
    --seed 42

Required Parameters:

• --model (or -m): HuggingFace repository or path to model checkpoint directory
• --prompt (or -p): Text prompt for video generation
• --erased-concept (or -e): Concept to erase from generation
• --output (or -o): Directory to save output video to

Optional Parameters:

• --seed (or -r): Random seed for reproducibility
• --width (or -W): Width of output video (default: 256)
• --height (or -H): Height of output video (default: 256)
• --num-frames (or -T): Total number of frames to generate (default: 16)

Supported Models

This implementation supports both ZeroScope and ModelScope backbones:

ZeroScope

# HuggingFace model ID
python inference.py \
    --model cerspense/zeroscope_v2_576w \
    --prompt "A man running under starry night by Van Gogh." \
    --erased-concept "Van Gogh" \
    --output ./outputs

# Or local path
python inference.py \
    --model models/zeroscope_v2_576w \
    --prompt "A man running under starry night by Van Gogh." \
    --erased-concept "Van Gogh" \
    --output ./outputs

ModelScope

# HuggingFace model ID
python inference.py \
    --model damo-vilab/text-to-video-ms-1.7b \
    --prompt "A man running under starry night by Van Gogh." \
    --erased-concept "Van Gogh" \
    --output ./outputs

# Or local path
python inference.py \
    --model models/modelscope \
    --prompt "A man running under starry night by Van Gogh." \
    --erased-concept "Van Gogh" \
    --output ./outputs