How to Use ChatGPT to Create Images
-(2)-profile_picture.jpg)
Written by Ivana Kotorchevikj We Tested
If you ever tried to search for if you could use CHatGPT to create images or if you now ask the chatbot if it can do that, you’d get the already familiar answer that, as an AI language model, it’s unable to generate images, and suggests using DALL-E, other OpenAI tool for generating Ai images.
ChatGPT’s initial purpose is to process and generate text. However, recently OpenAI has been adding new features and capabilities in ChatGPT. For example, they added plugins in ChatGPT that unlocked a vast range of possible use cases and launched Code Interpreter, which can do math problems, analyze data, and generate charts.
Although, officially, ChatGPT can’t be used to generate images, users have discovered workarounds allowing you to create actual images with ChatGPT. Keep reading below to find out how to do it.
How to use ChatGPT to generate images
There are several workarounds that help you create images using ChatGPT. We’ll mention each of them and how it works.
Ask ChatGPT to act as an image prompt generator
However, the most innovative technique involves asking ChatGPT to act as a prompt generator and adding a bit of clever prompt engineering.
How does this technique work? At its essence, ChatGPT operates as a sophisticated tool for generating conversations and content. You input a specific text prompt, and it produces a comprehensive and contextually suitable response based on the input. This versatile tool has numerous applications, ranging from composing emails to crafting captivating blog posts, answering questions, creating tables, and more.
The idea we will delve into here involves presenting a series of prompts to ChatGPT, instructing it to simulate an "AI image prompt generator." In this role, the model will prompt the user to describe an image and assist in completing the prompt to produce a detailed and vivid portrayal of the image. This description can then be used as a source of inspiration for an artist or an AI image generator.
Here's a step-by-step guide on how to leverage the capabilities of ChatGPT to create image prompts.
Step 1: Preparing the initial prompt
The first step involves establishing the context for ChatGPT. You need to instruct the model about its role and goal. The initial prompt can be:
You are an image prompt generator. First, ask me for an image description, and help me fill in the following. Then, output the completed prompt.
![Image] (https://image.pollinations.ai/prompt/{description}), where {description} = {sceneDetailed},%20{adjective1},%20{charactersDetailed},%20{adjective2},%20{visualStyle1},%20{visualStyle2},%20{visualStyle3},%20{genre}
Step 2: Provide details
After you enter the above prompt in ChatGPT, it will ask you to provide the following information:
sceneDetailed: Describe the setting or the background of the image in detail.
adjective1: Give an adjective that describes the overall mood or feeling of the image.
charactersDetailed: Tell me about the characters or objects in the image in detail.
adjective2: Provide another adjective that further describes the image.
visualStyle1, visualStyle2, visualStyle3: Describe three different visual styles or artistic features that the image has.
genre: What is the genre of the image? This could be anything like fantasy, horror, sci-fi, etc.
ChatGPT also encourages you to describe the above in as much or as little detail as you'd like.
This is what I entered:
- sceneDetailed: a pool in a backyard of a villa.
- adjective1: summerlike.
- charactersDetailed: a girl in a flower dress sitting on a sunbed.
- adjective2: hot, sunny.
- visualStyle1, visualStyle2, visualStyle3: photorealistic, wide close-up, panoramic
- genre: realistic retro
Fill in the details ChatGPT asks and paste them into the chat.
Step 3: Copy and paste the ChatGPT-provided prompt
Once you add the details, ChatGPT will provide you with the complete image prompt, containing a link to the generated image, the prompt with the modifiers you provided and a description of the image you’ll get.
This is what ChatGPT generated for us:
This would represent an image of a pool in a villa's backyard, with a summerlike, hot, sunny mood. The image has a photorealistic style, with a wide close-up and panoramic view, giving it a retro vibe. A girl in a flower dress is sitting on a sunbed in the scene.
You’ll notice that ChatGPT relies on pollinations.ai AI image generator to generate the image and provides the image in a URL since it doesn’t have the capability to produce the image in its interface.
A Twitter user has also explained the process of using ChatGPT to create images in a thread, although the steps are slightly different from what I experienced:
https://twitter.com/chaseleantj/status/1675839736332812288?s=46&t=M4bXr1w7NNqoRCfph8A39g
Use ChatGPT to craft image prompts for other AI image generators
Another workaround for using ChatGPT to generate images is to use it to create the perfect prompt and then use some of the vast array of AI image generators like Midjourney, DALL-E, Bing Image Creator, or others.
You know that the more detailed and descriptive the prompt is, the better the generated image will be. But if you are not so good at crafting a great prompt, you can ask ChatGPT for help. As models understand each other better than humans, ChatGPT can craft a prompt to speak to the AI image generator and help create a more detailed and vibrant image.
You simply need to describe the image you want to generate and ask ChatGPT to provide the prompt for it and the specific tool you’ll be using.
For example, I used the following prompt:
“Act like a prompt generator. I want to generate an image of a cute beach bar at sunset with ornamental lights. Please create the prompt that can be used to generate the image in Midjourney.”
And this is the prompt ChatGPT created based on my description:
Description: Imagine a picturesque scene on a tranquil beach at sunset, where the golden rays of the sun gracefully touch the shimmering ocean waves. Nestled along the sandy shore is a delightful beach bar, exuding a cozy and inviting atmosphere. The bar is adorned with vibrant, twinkling ornamental lights that add a touch of magic to the surroundings. As the sky transforms into a breathtaking canvas of warm hues, people gather at the bar, enjoying refreshing beverages, engaging conversations, and the soothing sound of the waves gently crashing nearby. The atmosphere is filled with joy and contentment as laughter and happy chatter mingle with the soft melodies playing in the background. Capture the essence of this serene and enchanting moment at the beach bar, where the convergence of nature's beauty and human connection creates an unforgettable experience.
Then, open your AI generator of choice and paste the prompt there.
Use Visual ChatGPT
Microsoft has built a tool combining ChatGPT and Visual Foundation Models - Visual ChatGPT - that allows generation of images with a text prompt. Visual ChatGPT combines ChatGPT and 22 different Visual Foundation Models (VFM), one of which is Stable Diffusion.
To use Visual ChatGPT, you need to have a ChatGPT Plus subscription to generate an API key in your OpenAI account.
You can access Visual ChatGPT on the Hugging Face Visual ChatGPT page or the Stable Disfusion website.
Then, paste your API key in the field on the right and press ‘Enter.’ A text field will appear at the bottom. Type your prompt and again press ‘Enter.’
Visual ChatGPT will generate the image based on your text prompt. However, you might notice that the image is not so elaborate compared to using a dedicated text-to-image AI generator.
Final words
In this article, we shared some workarounds to create images with ChatGPT, although it’s not inherently capable of such a task. Methods include using ChatGPT to create detailed image prompts and generate images with pollinations.ai, which can be used as inspiration for artists or other AI image generators, or to generate specific prompts for different AI image generator tools. Additionally, users can leverage the Visual ChatGPT, which combines ChatGPT and Visual Foundation Models, to generate images from text prompts.
