Step-by-Step Guide: Create Stunning AI-Generated Animated Reels
Here’s how the power of 👩🏻💻Stable Diffusion can help you make videos!
AI-Generated Art
Unless you live under a rock, you must have encountered some bizarre and extraordinary AI-generated Images. It’s incredible how technology is transforming the creative industry in such exciting ways! Computers can now learn to create art by analyzing large amounts of images and other media using advanced algorithms. They can generate various art forms like paintings, sketches, and animations. It’s like having your very own AI artist! Interestingly, this technology is helping artists and designers explore new styles and techniques previously impossible or difficult to achieve. And the best part is that AI-generated artwork has the potential to make art more accessible and affordable for everyone.
I have dabbled in creating AI-generated images in the past. You can check out my previous posts: https://avinashb.substack.com/p/heres-how-you-can-make-stunning-ai and https://avinashb.substack.com/p/effortlessly-create-stunning-ai-generated.
After Images, I wanted to create a Video because that’s where the magic happens. Now, the direct text-to-video output is currently in a primitive stage. I considered making a series of images and converting it into a GIF or a tiny video. And that’s what I did.
High-Level Plan
- Write a Text Prompt in Stable Diffusion to create an Image.
- Use the Image-to-Image function to generate similar images to the source image.
- Use a Video Editor to convert this set of images into a movie.
Step 1 — Text Prompt
I use Automatic 1111 to run Stable Diffusion locally on my PC. I have a beefy MSI GeForce RTX 3060 Ti GAMING X TRIO that can handle any game on the market and pull its weight in Image rendering. Now the output image style depends on the Stable Diffusion Model used for processing. I used the Chillout Mix Model from Civitai for this Project.
Selecting a Model is an integral part of the Image Generation process. But do you know what’s more important? The text prompt, of course. When using an AI model to create an image, you start by giving it a text prompt, a short description of what you want the Image to look like. This can be something specific, like “a yellow sunflower with a blue background” or, more generally, “a peaceful forest scene.” The text prompt is essential because it gives the AI model a starting point to work from and helps it understand what you’re looking for. Without a good text prompt, the AI model might generate something completely different from what you had in mind! That’s why it’s essential to be as clear and detailed as possible when writing your text prompt. The more specific you are, the better your chance of getting the Image you want.
I asked Chat GPT to come up with an elaborate and detailed prompt for Image Generation based on references from Civitai, and here’s what it threw up at me-
(Dawn photo, pastel colors), (realistic, photo-realistic:1.3), ethereal, extremely delicate and beautiful, extremely detailed, CG, unity, 2k wallpaper, Exquisite, finely detailed, serene smile, extremely detailed CG unity 8k wallpaper, huge filesize, ultra-detailed, high res, absurd res, gentle light, (((medium hair:1.3), loose, black hair, floral hairpin)), beautiful detailed girl, detailed fingers, extremely detailed eyes and face, beautifully detailed nose, large beautiful detailed eyes, long eyelashes, looking at the viewer, (closed mouth:1.2), 1girl, pretty, young, mature face, (upper body:1.3), realistic face, realistic body, flowing Anarkali suit, delicate jewelry, simple mehndi, barefoot.
Once I had the prompt, I executed this to generate around 100 Images. With these Images, I shortlisted a few Images for the next step.
Step 2— Image to Image Function
Automatic 1111 has an Image to Image Tab that can generate multiple Images based on a Source Image. I selected my shortlisted Image and generated a batch of 80 Images. I kept all Settings the same as what was set during Text to Image Creation. Denoising Strength was between 0.45 to 0.65 during different Images. Denoising Strength determines how identical or different you want the Batch of Images to be from the original Shorlisted Image. 0 means an exact Image, and 1 represents an entirely different Image. I repeated this process for all of my shortlisted images.
Step 3 — Create a Movie
I have been using DaVinci Resolve for all my Video Editing Needs since it works well on Windows and is free. I imported the Batch of Photos onto DaVinci Resolve and exported the movie. I selected an appropriate Vertical Resolution of 1080x1920 to make the output video look good on Instagram Reels.
Conclusion
It took a few hours to complete the process, and I was impressed by the outcome. Do you have any recommendations or ideas that you want me to try out? Please let me know in the comments below.
