clipyr / README.md
Gresekxnol's picture
Update README.md
b46e963 verified

A newer version of the Gradio SDK is available: 5.44.1

Upgrade
metadata
title: AI Video Clipper
emoji: 🎬
colorFrom: purple
colorTo: pink
sdk: gradio
sdk_version: 5.42.0
app_file: app.py
pinned: false
license: mit
short_description: Transform long videos into viral short clips using AI
tags:
  - video-editing
  - ai
  - social-media
  - clips
  - tiktok
  - instagram
  - youtube-shorts
  - content-creation
  - viral
  - auto-captioning

🎬 AI Video Clipper

Transform your long-form videos into viral short clips automatically using AI! This application mimics the core functionality of OpusClip with open-source technologies.

✨ Features

πŸ€– AI-Powered Analysis

  • Smart Moment Detection: Uses advanced NLP to identify the most engaging parts of your video
  • Virality Scoring: Predicts which clips have the highest potential for social media success
  • Multi-Modal Analysis: Combines audio features, sentiment analysis, and content understanding

πŸ“± Social Media Optimization

  • Auto 9:16 Conversion: Automatically converts to vertical format perfect for TikTok, Instagram Reels, YouTube Shorts
  • Smart Cropping: Intelligently crops and scales video content
  • Background Blur: Creates cinematic background when needed

πŸ“ Auto-Captioning

  • High-Accuracy Transcription: Uses OpenAI Whisper for precise speech-to-text
  • Emoji Enhancement: Automatically adds relevant emojis to increase engagement
  • Multi-Language Support: Works with videos in multiple languages

🎯 Content Intelligence

  • Hook Detection: Identifies compelling opening statements and cliffhangers
  • Emotional Analysis: Detects high-engagement emotional moments
  • Keyword Optimization: Recognizes viral keywords and phrases

πŸš€ How It Works

  1. Upload: Drop your video file (MP4, AVI, MOV, MKV, WebM)
  2. Configure: Set clip duration (15-90 seconds) and number of clips (1-5)
  3. Process: AI analyzes your video for the best moments
  4. Download: Get optimized clips ready for social media

πŸ“Š Virality Scoring Algorithm

The AI evaluates clips based on:

  • Sentiment Analysis: Positive/negative emotional impact
  • Emotional Intensity: Surprise, excitement, engagement levels
  • Viral Keywords: Money, secrets, transformations, reactions
  • Hook Patterns: "You won't believe", "This will change", etc.
  • Audio Features: Tempo, energy, vocal dynamics
  • Optimal Length: 30-60 second sweet spot for maximum retention

πŸŽ₯ Best Practices

Ideal Content Types:

  • Podcasts & Interviews: Great for extracting key insights and memorable quotes
  • Educational Content: Tutorial highlights and "aha moments"
  • Webinars & Presentations: Key takeaways and surprising statistics
  • Product Reviews: Before/after comparisons and reactions
  • Storytelling: Climactic moments and emotional peaks

Tips for Best Results:

  • βœ… Upload videos with clear, audible speech
  • βœ… Longer videos (5+ minutes) provide more clip opportunities
  • βœ… Content with emotional moments scores higher
  • βœ… Good lighting and stable footage improve final quality
  • ❌ Avoid videos with background music louder than speech
  • ❌ Very quiet or poor audio quality may affect transcription

πŸ”§ Technical Details

AI Models Used:

  • Whisper Base: For speech transcription and timestamp alignment
  • RoBERTa Sentiment: For emotional analysis and engagement prediction
  • DistilRoBERTa Emotion: For detecting surprise, excitement, and other high-engagement emotions

Processing Pipeline:

  1. Audio Extraction: Separates audio track for analysis
  2. Transcription: Generates timestamped text with word-level precision
  3. Feature Analysis: Extracts tempo, energy, and acoustic features
  4. Content Scoring: Evaluates each segment for viral potential
  5. Clip Generation: Creates optimized vertical videos with subtitles
  6. Quality Enhancement: Applies blur backgrounds and text styling

πŸ“ˆ Performance Optimizations

  • Efficient Processing: Uses base models optimized for speed
  • Smart Batching: Processes multiple segments simultaneously
  • Memory Management: Automatic cleanup prevents memory leaks
  • Fast Encoding: Uses ultrafast presets for quick turnaround

🎨 Customization Options

  • Clip Duration: 15-90 seconds (30-60s recommended)
  • Number of Clips: 1-5 clips per video
  • Subtitle Toggle: Enable/disable auto-generated captions
  • Emoji Integration: Automatic contextual emoji insertion

πŸ”’ Privacy & Security

  • No Data Storage: Videos are processed temporarily and automatically deleted
  • Local Processing: All AI analysis happens on the server without external API calls
  • Secure Upload: Files are handled securely with temporary storage only

πŸš€ Getting Started

This application is designed to run on Hugging Face Spaces with the free tier. Simply upload your video and let the AI work its magic!

Perfect for:

  • πŸ“± Content creators looking to repurpose long-form content
  • 🎬 Marketers creating social media campaigns
  • πŸŽ“ Educators extracting key teaching moments
  • πŸ’Ό Businesses showcasing product highlights
  • πŸŽ™οΈ Podcasters creating promotional clips

Built with ❀️ using open-source AI technologies. Transform your content strategy today!