Canstralian/ai-video-composer

🏞 AI Video Composer

AI Video Composer is an intelligent media processing application that uses natural language instructions to create videos from your media assets. It leverages the Qwen2.5-Coder language model to generate FFmpeg commands based on your requirements.

How It Works

1. 1.Upload Media Files:

• —Supports multiple file formats including:
• —Images: .png, .jpg, .jpeg, .tiff, .bmp, .gif, .svg
• —Audio: .mp3, .wav, .ogg
• —Video: .mp4, .avi, .mov, .mkv, .flv, .wmv, .webm, and more
• —File size limit: 10MB per file
• —Video duration limit: 2 minutes

1. 1.Provide Instructions:

• —Write natural language instructions describing how you want to process your media
• —Examples:
• —"Convert these images into a slideshow with 1 second per image"
• —"Add this audio track to the video"
• —"Make the video play 2x faster"
• —"Create a waveform visualization for this audio file"

1. 1.Advanced Parameters:

• —Top-p (nucleus sampling): Controls diversity of generated commands (0-1)
• —Temperature: Controls randomness in command generation (0-5)

1. 1.Processing:
2. 2.The app analyzes your files and instructions
3. 3.Generates an optimized FFmpeg command using Qwen2.5-Coder
4. 4.Executes the command and returns the processed video
5. 5.Displays the generated FFmpeg command for transparency

Features

• —Smart Command Generation: Automatically generates optimal FFmpeg commands based on natural language input
• —Error Handling: Validates commands before execution and retries with alternative approaches if needed
• —Multiple Asset Support: Process multiple media files in a single operation
• —Waveform Visualization: Special support for audio visualization with customizable parameters
• —Image Sequence Processing: Efficient handling of image sequences for slideshow creation
• —Format Conversion: Support for various input/output format conversions
• —Example Gallery: Built-in examples demonstrating common use cases

Technical Details

• —Built with Gradio for the user interface
• —Uses FFmpeg for media processing
• —Powered by Qwen2.5-Coder for command generation
• —Implements robust error handling and command validation
• —Processes files in a temporary directory for safety
• —Supports both simple operations and complex media transformations

Limitations

• —Maximum file size: 10MB per file
• —Maximum video duration: 2 minutes
• —Output format: Always MP4
• —Processing time may vary based on input complexity

Contributing

If you have ideas for improvements or bug fixes, please open a PR: