Summary

ViralCutter is a free, open-source desktop application designed to transform long YouTube videos into short, viral clips optimized for platforms like TikTok and Instagram Reels. It addresses the high costs and data privacy concerns associated with commercial SaaS alternatives by offering unlimited, local video processing with advanced AI features.

How It Works

The project leverages state-of-the-art AI models for content analysis, including options for cloud-based Gemini/GPT-4 or local GGUF models (Llama 3, DeepSeek) for identifying engaging segments. It utilizes WhisperX with GPU acceleration for highly accurate transcriptions, which are then styled into dynamic, word-highlighted captions. Core features include automatic 9:16 aspect ratio cropping, intelligent split-screen detection for dual speakers, and experimental active speaker tracking, all processed locally for enhanced user privacy.

Quick Start & Requirements

Installation involves running either install_dependencies.bat (standard, uses online AI) or install_dependencies_advanced_LocalLLM.bat (for offline AI) within the project directory. These scripts utilize the uv package manager. Prerequisites:

• Build Tools: Visual Studio Build Tools with "Desktop development with C++" workload (including Windows 10/11 SDK and MSVC v143).
• Python: Version 3.10.x or 3.11.x recommended, with "Add Python to PATH" enabled during installation.
• FFmpeg: Installable via winget install ffmpeg in PowerShell (as Administrator).
• GPU: An NVIDIA GPU with CUDA 12.4+ support is strongly recommended for optimal performance and local AI inference. Running: Execute run_webui.bat to launch the Gradio-based web interface or python main_improved.py for the CLI. API keys for cloud AI services are configured in api_config.json, while local models (.gguf) are placed in the models/ directory.

Highlighted Details

• AI Viral Clipping: Automatic identification of engaging video segments using flexible AI backends.
• Transcription & Captions: Ultra-accurate, GPU-accelerated transcription via WhisperX, with highly customizable, dynamic "Hormozi"-style captions.
• Automated Editing: Intelligent 9:16 auto-cropping, split-screen for two speakers, and experimental active speaker focus.
• Multilingual Support: Automatic translation of generated subtitles into over 10 languages.
• Professional Output: Supports up to 4K resolution and includes a Beta feature for exporting XML files compatible with Adobe Premiere Pro.

Maintenance & Community

ViralCutter is community-maintained and actively seeking contributions. The project is currently in an Alpha stage (v0.8v Alpha). Community interaction is facilitated via the "AI Hub Brasil" Discord server. Users are encouraged to star the GitHub repository.

Licensing & Compatibility

The project is described as "Free" and "Open-Source," but a specific software license (e.g., MIT, GPL) is not explicitly stated in the provided README. This lack of explicit licensing may pose compatibility questions for commercial use or integration into proprietary software.

Limitations & Caveats

As an Alpha release, ViralCutter may contain bugs or incomplete features. The "Active Speaker" tracking functionality is noted as experimental. Running local LLMs requires substantial GPU VRAM and processing power. Several planned features, such as automatic background music ducking and direct platform uploads, are still under development. The absence of a formal license requires careful consideration for any form of redistribution or commercial application.