AutoSlice

AutoSlice is a localized workflow tool specifically designed for Bilibili recordings, aiming to transform long video recordings into high-quality short video slices through automation. The tool integrates speech-to-text, danmaku analysis, and Large Language Models (LLM), providing end-to-end processing capabilities from content analysis to the final output. Key features include: 1. **Intelligent Analysis and Slicing**: Utilizes FunASR for speech-to-subtitle conversion, combining ASS/XML danmaku peaks and optional manual timelines (DOCX) to discover highlight topics, generate topic reports, and filter slice candidates. 2. **Subtitle Processing**: Supports subtitle generation using FunASR models (such as Fun-ASR-Nano-2512) and provides a subtitle proofreading workbench, allowing users to burn-in hard subtitles to MP4 after confirming AI suggestions. 3. **Cover Production**: Built-in AutoCover linkage functionality supports selecting frames from slices and adjusting titles, stickers, and composition, with the ability to save local editing drafts. 4. **Media Processing**: Supports various video formats (.flv, .mp4, .mkv, .mov, .avi), implements precise slicing via FFmpeg, and can connect to CapCut (Jianying) for fine-tuning. The tool is primarily aimed at Windows 10/11 platforms, supports optional NVIDIA GPU acceleration, and provides a Flask-based Web interface for operation and management.