How to Make Ghibli Style Images for Free

Ever wondered how to create stunning Ghibli-style images, or even better, for free? With the rise of AI image generators, it's now easier than ever to bring your creative visions to life. In this guide, we'll explore tools to generate beautiful images that capture the essence of Studio Ghibli's enchanting art style.

What is Studio Ghibli Art Style?

Studio Ghibli is a renowned Japanese animation studio known for its breathtaking visuals and captivating storytelling. The Ghibli art style is characterized by lush landscapes, whimsical characters, and a sense of wonder that transports viewers to magical worlds. To create images in this style, we need to focus on elements like color palettes, composition, and character design.

Ways to Create Ghibli Art Style Images

There are several ways to create Ghibli-style images, each with its own advantages and disadvantages. Here are some of the most popular methods:

Our recommandation:

• ChatGPT: If you're just looking for a quick and easy way to generate Ghibli-style images, ChatGPT is a great option.
• Hosted Models: If you're a developer or want more control over the image generation process, using hosted models is a good choice.
• Free AI Image Generators: If you're AI enthusiast and want to experiment with different models, free AI image generators are a great way to get started.
• Local AI Image Generators: If you're serious about creating Ghibli-style images and have access to a powerful GPU, using local AI image generators is the best option.

Use GPT-4o in ChatGPT

The new GPT-4o model from OpenAI can generate images based on text prompts. You can describe the scene you want to create, and the model will generate an image that matches your description. This also works if you prompt for images to be "in Ghibli art style".

To generate Ghibli-style images using ChatGPT, follow these steps:

1. Log into ChatGPT website or mobile app. Or if you don't have an account, you can sign up for free.
2. Chose the GPT-4o model. Free users may not have the option but default to use the GPT-4o model, while paid users can select the GPT-4o model from the model selection menu.
3. Type in your prompt. For example, "Create a Ghibli-style image of a magical forest with glowing creatures.". Make sure to include "Ghibli style" in your prompt to ensure the model understands the desired art style.
4. Submit your prompt and wait for the model to create your image. It would take a few seconds to a minute for the model to generate the image.
5. Once the image is generated, you can click on it to view it in full size. You can also download the image by right-clicking on it and selecting "Save image as...". Further editing can be done by following prompts like "Add a castle in the background" or "Make the forest more colorful".

What we love about using ChatGPT for image generation:

• User-friendly: The chat interface is easy to navigate, providing both web and mobile access.
• High-quality images: The images generated are of high quality, and the model is capable of producing stunning Ghibli-style images.

What we don't love:

• Limited access: The free version of ChatGPT has limitations of 4 images per day. You need to wait in queue for image to be created. This can be frustrating if you're trying to create multiple images.
• Strict restrictions: OpenAI has strict policies regarding the use of copyrighted material, which can sometimes refuse to generate the images in desired art styles.

Who is it for?

• You just want to create Ghibli-style images quickly and easily.
• You prefer chatting like interaction instead of tweaking a lot of parameters.

Use Hosted Models

While it still costs a little bit, it may be the best way if you want balance between user-friendliness and less restrictions. Some of the open weights models like FLUX, Stable Diffusion and others models can offer Ghibli-style models, but requires complex workflows to work well. Those hosted models can greatly simplify the process of generating Ghibli-style images. They provide a user-friendly interface and handle all the technical details for you, allowing you to focus on your creative vision.

Some notable hosted models include:

• The EasyControl Ghibli app on Hugging Face Spaces. This is probably the best option to try. Hugging Face is the leading platform for AI models and has a large community of developers and enthusiasts. You may find the UI of EasyControl Ghibli app a little bit strange, but it is actually get the job done.
• The Ghibli model on Replicate. Replicate is popular among AI developers and researchers. It provides a community for hosted models that you can try on the web, or use the API to integrate into your own applications.

What we love about using hosted models for image generation:

• No restrictions. You can generate images without worrying about copyright issues or other restrictions.
• Free or low cost. Most hosted models are free to use, or have a low cost for generating images.

What we don't love:

• Usability. Most of those hosting platforms are not as user-friendly as ChatGPT. It may take some time to get used to the interface and features.

Who is it for?

• You want to create Ghibli-style images without constraints.
• You want to integrate the image generation process into your own applications.

Use Ghibli model on Hugging Face Spaces

Go to the EasyControl Ghibli app on Hugging Face Spaces. Click to upload an image or drag and drop an image to the app. You can also use the default image provided by the app. Once the image is uploaded, you can adjust the parameters to create your Ghibli-style image.

Some parameters you can adjust include:

• Prompt. The default prompt usually works well, so you can leave it as is at first. If you want to improve the image, keep the default prompt and prepend your prompt to the default prompt.
• Height and width of the generated image. The maximum size is 1024x1024 pixels. And there's one limitation that the image size must be divisible by 8. For best output images, you may want to set the dimensions of the image to the aspect ratio of the original image.
• Use CFG zero star and zero init steps. Checking this option and set zero init steps to 1 usually results in better images. Those are basically improvements to the controllablity of the model.
• Seed and control type. Those are not necessary to change.

Once you're satisfied with the parameters, click the "Generate image" button to create your Ghibli-style image. The app will take a few seconds to generate the image, and once it's done, you can download it by clicking on the image.

Hugging Face Spaces also puts quotas on GPU resouces you can use per day. If you want to generate more images, you may have to sign up for a paid plan. But generally, it has less restrictions than ChatGPT.

Use Ghibli model on Replicate

Go to the Ghibli model on Replicate. Click to upload an image or drag and drop an image to the app in the spatial_img field. You can also click "tweak it" on the examples to use the input image with prefilled parameters.

Some parameters you can adjust include:

• Prompt. The default prompt usually works well, so you can leave it as is at first. If you want to improve the image, keep the default prompt and prepend your prompt to the default prompt.
• Height and width of the generated image. The maximum size is 1024x1024 pixels. And there's one limitation that the image size must be divisible by 8. For best output images, you may want to set the dimensions of the image to the aspect ratio of the original image.
• Use CFG zero star and zero init steps. Checking this option and set zero init steps to 1 usually results in better images. Those are basically improvements to the controllablity of the model.
• Seed and control type. Those are not necessary to change.

Once you're satisfied with the parameters, click the "Run" button to generate your Ghibli-style image. The app will take a few seconds to generate the image, and once it's done, you can download it by clicking on the image.

Technically it's not free to generate images on Replicate. And you are required to add payment method to your account before you can generate images. The cost is about $0.01 per image for GPU computes.

Use Free AI Image Generators

There are several free AI image generators available online that can help you create Ghibli-style images. Some of the most popular ones include:

• PixelsAI. We at PixelsAI have created an Ghibli-style image generator with free trial. It's user-friendly and privacy-focused.
• Ghiblio.art. This is a Ghibli art generator that provides free trial. It also support other styles.

What we love about using free AI image generators:

• Free trial. Most free AI image generators offer a free trial, allowing you to test the service before committing to a paid plan.
• User-friendly. Free AI image generators are usually easy to use, with a simple interface that allows you to create images quickly.

What we don't love:

• Image quality varies. As the time of writting, the image quality of free AI image generators is not as good as the paid ones. The images may not be as detailed or realistic as those generated by paid services.

Who is it for?

• You want to experiment different services of Ghibli-style image generators for free.

Use Local AI Image Generators

When you have access to a powerful GPU, using local AI image generators is the great option. ComfyUI is a powerful and popular tool to create complex image workflows. It may seem intimidating at first, but once you get the hang of it, you'll find it to be a powerful tool for creating Ghibli-style images, especially with a large community.

For how to install ComfyUI, you can download the desktop version from its website here. Or if you are tech savvy, you can install it from the source code on GitHub repo.

Once your have ComfyUI installed and running, try the following models and workflows to create Ghibli-style images.

• Ghibli Art Style High Quality. It's a LoRA model that can be used with Stable Diffusion XL. Make sure to include the trigger word "Ghibli" in your prompt to get the best results.
• Studio Ghibli Style LoRA. This is also a LoRA model that works with Stable Diffusion 1.5. The quality is pretty solid despit the base model is not as good as Stable Diffusion XL.

It's also possible to generate Ghibli-style images without ComfyUI. One of our favorite is the Ghibli model in EasyControl project. You need basic knowledge of Python and Git to run it. The model is based on FLUX.1-dev and can generate Ghibli-style images with high quality. Following instructions in the project README to load the model and run it.

What we love about using local AI image generators:

• No restrictions. You can create Ghibli-style images without worrying about copyright issues or other restrictions.
• Privacy. You have full control over your images and data, ensuring your privacy is protected.
• High-quality images. There are high quality models available.

What we don't love:

• Requires powerful GPU. Local AI image generators require a powerful GPU to run, which may not be accessible to everyone.
• Complex setup. Setting up local AI image generators can be complicated and time-consuming, especially for beginners.

Who is it for?

• You have access to a powerful GPU and want to create Ghibli-style images without restrictions.
• You are comfortable with technical setups and want to experiment with different models and workflows.

Conclusion

Creating Ghibli-style images has never been easier, thanks to the advancements in AI image generation technology. Whether you choose to use ChatGPT, hosted models, free AI image generators, or local AI image generators, there are plenty of options available to suit your needs and preferences.