video-subtitle-remover

Project Overview

Video-subtitle-remover (VSR) is an AI-based software for removing hard-coded subtitles from videos. It mainly achieves: removing hard-coded subtitles from videos at lossless resolution, generating subtitle-free files; using powerful AI algorithm models to inpaint the removed subtitle text regions (not adjacent pixel filling or mosaic removal); extracting original video subtitles, usable with video-subtitle-extractor (VSE).

Key Features

• Remove hard-coded subtitles at lossless resolution, generating subtitle-free files
• Inpaint removed subtitle text regions using powerful AI algorithm models
• Support custom subtitle positions, removing only subtitles in defined locations
• Support automatic removal of all text across the entire video
• Support batch removal of watermark text from multiple selected images

Target Audience

• Users who need to remove hard-coded subtitles from videos
• Users who need to remove watermarks from images/videos
• Video editors and content creators

Quick Start

1. Download the compressed package and extract to run
2. If it doesn't run, try installing a conda environment from source following the tutorial
3. Download address: Release page

Core Advantages

• AI-based, implemented locally without requiring third-party APIs
• Lossless resolution removal of hard-coded subtitles and text watermarks
• Supports custom subtitle positions and automatic full-video removal
• Provides Windows pre-built packages (CPU, DirectML, CUDA variants) and Docker versions

System Requirements

• OS: Windows (pre-built packages), Linux / macOS (Docker)
• Additional dependencies: Depends on the version; NVIDIA GPUs require the corresponding CUDA version; AMD/Intel GPUs use the DirectML version