How to make GIF Animations with Stable Diffusion (AnimateDiff)

Updated
How to make GIF Animations with Stable Diffusion (AnimateDiff)
Create stunning GIF animations with AnimateDiff! Learn how to use this powerful tool for Stable Diffusion and unleash your creativity.

1. Introduction to AnimateDiff

In this tutorial I'm going to show you AnimateDiff, a tool that allows you to create amazing GIF animations with Stable Diffusion. The best text to video AI tool available right now.

The AnimateDiff GitHub page is a source where you can find a lot of information and examples of how the animations are supposed to look. For example, you can create realistic videos with it, or a cartoon style, at the moment the cartoon style works the best, so in this tutorial I will be showcasing on how to use that.

2. Downloading the Requirements

A good checkpoint for this cartoon style is ToonYou, you can download the latest version from the CivitAI website and install it in your Stable Diffusion models folder. Download ToonYou from the official CivitAI website.

how-to-make-an-ai-generated-gif-toonyou.webp Place this checkpoint in your Stable Diffusion models folder as shown below. how-to-make-an-ai-generated-gif-toonyou-install.PNG

Stable Diffusion in the Cloud⚡️Run Automatic1111 in your browser in under 90 seconds
20% bonus on first deposit

3. Installing AnimateDiff Extension

To get started, you don't need to download anything from the GitHub page. Instead, go to your Stable Diffusion extensions tab. Click on "Available", then "Load from", and search for "AnimateDiff" in the list. Click on "Install" to add the extension. If you can't find it in the search, make sure to Uncheck "Hide Extensions with tags -> Script" and it will appear.

how-to-make-an-ai-generated-gif-install-extension.PNG

After the extension has been installed go to the “Installed” tab and click on “Apply and restart UI”. I also recommend restarting your Stable Diffusion completely to prevent any errors from occuring.

how-to-make-an-ai-generated-gif-restart-ui.PNG

4. Installing AnimateDiff Model

Next we need the AnimateDiff model, this is required to use the extension, you can download this model from the Hugging Face website. Currently there are 2 models available; “mm_sd_v14.ckpt” & “mm_sd_v15.ckpt”. I’ve found that the v14 model works way better but I recommend downloading both so you can try them out yourselves.

Place these files in the following directory: "Stable Diffusion"> "Extensions"> "SD web UI animatediff"> "models".

how-to-make-an-ai-generated-gif-model.PNG

Stable Diffusion in the Cloud⚡️Run Automatic1111 in your browser in under 90 seconds

5. AnimateDiff Settings

Once you've installed the AnimateDiff extension, it will be available at the bottom of the Stable Diffusion interface. To use it, click on the "AnimatedDiff" option and the interface should fold out. Here are some settings you can configure, I will list the settings I recommend.

  • Motion module: mm_sd_v14.ckpt
  • Number of frames: I recommend at least 8 frames for good quality, if you use a lower value the output won’t be as nice.
  • Frames per second (FPS): Adjust the playback speed, I recommend at least 8 to 12.

Don't forget to check "enable" to use the extension. how-to-make-an-ai-generated-gif-animatediff-settings.PNG

Troubleshooting

If you experience a long generating time press "Remove motion module from any memory" before generating. It's also helpful to keep the negative prompt under 75 characters. Use max 16 number of frames.

6. Stable Diffusion Settings

I will list the recommended settings for Stable Diffusion with the ToonYou checkpoint. You can experiment with these settings and find out what works best for you.

  • Checkpoint: ToonYou
  • Clip skip: 2 (or higher)
  • Prompt: Include “(best quality, masterpiece)” for better results
  • Negative Prompt: Include “(worst quality, low quality, letterboxed)” for better results
  • Sampling method: DDIM (This is by far the fastest sampling method and will decrease generating time by a lot).
  • Sampling steps: A minimum of 25, I recommend 40.
  • Width & Height: 512x512 or 768x768. Remember you can always upscale your GIF later.
  • CFG Scale: 7.5 - 8

Just stick with the default settings for the rest of the configurations. how-to-make-an-ai-generated-gif-all-settings.PNG

What I recommend before generating your GIF is to generate some images and use the seed from the image you liked the best.

Now you're ready to generate amazing animated GIFs powered by AI!

Stable Diffusion in the Cloud⚡️Run Automatic1111 in your browser in under 90 seconds

7. Examples

Here are some examples of GIF animations rendered using AnimateDiff with specific settings:

Example 1

  • Sampling method: DDIM
  • Steps: 40
  • Resolution: 768x512
  • CFG Scale: 8
  • Model: ToonYou (Beta 6)

how-to-make-gif-animations-with-stable-diffusion-animatediff-gif.gif

Example 2

  • Sampling method: DDIM
  • Steps: 40
  • Resolution: 512x512
  • CFG Scale: 8
  • Model: ToonYou (Beta 6)

how-to-make-an-ai-generated-gif.gif

8. Conclusion

While the quality of the output might vary, AnimateDiff is still a powerful tool that can produce amazing animations. Try experimenting with different settings and models to achieve the best results for your projects.

Frequently Asked Questions

AnimateDiff is primarily designed for use with Stable Diffusion, but it can also be utilized in other environments like Google Colab with some adjustments.

AnimateDiff should work on most modern computers with Stable Diffusion installed. Ensure that you have sufficient RAM and GPU capabilities for optimal performance.

To get the best results, use the recommended settings for Stable Diffusion, including the ToonYou checkpoint, a minimum of 30 steps, and a resolution of 512x512 or higher.

Yes, you can experiment with various text prompts to generate different animation outcomes. Feel free to get creative and see what works best for your desired results.

AnimateDiff can be used by users with different levels of experience. Beginners can start with default settings and gradually explore more advanced configurations as they gain familiarity with the tool.