video-automation

Project Overview

Video Automation is an AI clipping tool that runs on your own computer, converting long local videos (livestream replays, recordings) into short videos for Douyin, Bilibili and YouTube Shorts: smart segment selection, transcript subtitles, cover images, and exported finished files.

It ships as a Windows desktop app and can also run from Python source; transcription uses a local Faster-Whisper model by default, the basic flow needs no API key, and an external AI service or an NVIDIA GPU can optionally speed things up.

Item	Value
Platform	Windows (app), macOS and Linux (from source)
Language	Python 3.11+
Transcription	Faster-Whisper / FunASR (local)
Dependencies	FFmpeg, FFprobe
License	MIT

Long video in, shorts out, with subtitles and covers handled end to end, all processed locally.

Key Features

• Smart clipping: automatically picks valuable segments from long videos into reviewable draft shorts
• Auto subtitles: local speech-to-text subtitle generation with Faster-Whisper and FunASR engines
• AI covers: generates candidate cover images to pick from and composite into the final cut
• Platform presets: built-in Fast, Douyin and Bilibili profiles matched to each platform with one click
• Batch processing: add up to 30 files at once from the recording list, duplicates skipped
• Damage repair: re-cut corrupted parts via Exclude damaged time range, with undo support
• Dependency self-check: the Health page detects FFmpeg and FFprobe and fixes missing ones automatically
• Review workspace: review and edit kept clip text in the finished job, then export final.mp4

Target Audience

• Streamers turning replays into shorts
• Creators batch-processing recorded shows
• Privacy-minded creators who avoid cloud uploads
• Developers who want to customize the pipeline in Python

Quick Start

1. Download the Windows package from Releases, install and run VideoAutomationLite.exe
2. Open the Health page; if FFmpeg is missing click Auto-fix Dependencies
3. Add a local video in New Job, or use Add all to batch for many files
4. Pick the Fast, Douyin or Bilibili profile and start processing
5. Open the finished job, review it, and download final.mp4

Run from Source

1. Prepare Python 3.11+, FFmpeg, FFprobe and Git
2. Clone the repo and create a virtual environment
3. Install dependencies pip install -r requirements-transcription-faster.txt
4. Start the service with python run_worker.py --serve

Core Advantages

• Fully local: transcription, clipping and covers all run on your machine, originals never modified
• Zero-config output: the basic flow needs no API key, ready right after install
• Acceleratable: an NVIDIA GPU noticeably speeds up transcription and rendering
• Clear boundaries: no URL downloading or livestream recording, platform login and auto-publish disabled by default

System Requirements

Item	Requirement
OS	Windows 10/11 (app); macOS, Linux (source)
Runtime	Python 3.11+, FFmpeg, FFprobe
Hardware	Optional NVIDIA GPU for faster transcription and rendering

Notes

• Local files only: URL downloading and livestream recording are not included, prepare local videos first
• Timestamp warnings: a warning alone is not proof of damaged frames, confirm before re-cutting with exclusion
• Deletion and disk space: batch-deleted files move to .deleted_recordings, clean manually to free space