Nova Reel Prompt Optimizer

A Gradio web interface for optimizing prompts and generating videos using Amazon's Nova Canvas and Reel model.

Update Logs

12/30/2024

• Add Long Video Generation.
  Expand an input from the user, generate a storyboard, create breakdown shots, and generate a long video

• 1. Prompt:我心目中的AWS像一柄利剑，帮我在数字化转型的道路上披荆斩棘。 video demo1 file

• 

• 2. Prompt:wonderful under sea world video demo2 file

• 

example:

1/6/2025

• Add QR Code generation for long video. The viewers can download the artifacts in their phone

12/23/2024

• Add seed, cfg scale, Aspect ratio
• Add option to enable/disable comparison mode

Features

• Text prompt optimization for Nova Canvas image generation
• Text prompt optimization for Nova Reel video generation
• 1-click automatically optimize any languages & any prompts, with the reference of Nova official guideline.
• 1-click to copy the generated image from Canvas to for Reel video generation
• Real-time prompt optimization using Nova
• Comparison Videos generation from original prompt and optimized prompt
• User-friendly web interface

Prerequisites

• Python 3.10+
• AWS credentials configured with access to Bedrock and S3
• An S3 bucket for video output (You can copy the default bucket created from Nova Reel when you first time play with it in Bedrock Console)

Installation

1. Install the required Python packages:

pip install -r requirements.txt

2. Make sure you have AWS credentials configured with appropriate permissions for:

   • Amazon Nova Pro/Lite/Canvas and Reel
   • S3 bucket access

3. Update the S3 bucket name in WebUI or you can change it in config.py DEFAULT_BUCKET=:

Usage

1. Start the Gradio interface:

python app.py

2. Open your web browser and navigate to the URL shown in the terminal (typically http://127.0.0.1:7860)

3. Image generation

   • Enter your prompt in the text box
   • Click "Optimze Prompt" to start the create Optimized Prompt and Negative Prompt (If needed)
   • Click "Generate Image" to start the create images
   • View the optimized prompt and generated images
   • 1-click to copy the image to Video generation

4. Video generation

   • Enter your prompt in the text box
   • Optionally upload an image or 1-click copy from Image generation
   • Click "Optimze Prompt" to start the create Optimized Prompt and Negative Prompt (If needed)
   • Click "Generate Image" to start the create image
   • View the optimized prompt and generated video
   • You can compare the videos between original and optimized version

5. Long video generation

   • Enter your prompt in the story box
   • Click "Generate Shots" to create storyboard.
   • Click "Generate Video" to generate a long video.
     It first creates images by scens, generate videos from every image of the scen, then concate the short videos to a long video, add caption to the video s

Notes

• The application uses Amazon Nova to optimize prompts for better video generation results
• I found sometimes that Nova Pro works better for image optimzation while Nova Lite works better for video, so you can switch the model to try.
• Video generation may take several minutes to complete
• Generated videos are temporarily stored in the generated_videos directory
• Generated images are temporarily stored in the generated_images directory
• Both English and Chinese prompts are supported, but optimized prompts will be in English

Example Prompts

Text to Image example:

海底漫游的视频

a video of walking in a underwater world