Can ChatGPT generate images? Yes, and here's how to do it

Since ChatGPT’s release in late 2022, the chatbot has received numerous updates to add features like web browsing support and third-party plugins. Not to mention, the latest GPT-4 Turbo model adds multi-modal capabilities that allow the chatbot to go beyond the written word with vision and speech. With that in mind, you may wonder: can ChatGPT generate images like other AI platforms? The answer’s a bit complicated, so let’s break it down.

Can ChatGPT generate images?

Yes, ChatGPT can generate images but it relies on DALL-E 3 for all of the heavy lifting. San Francisco-based startup OpenAI created both ChatGPT and DALL-E, so it’s easy to see why the two AI tools are interdependent. It didn’t always work this way, though — for roughly a year, you could only use ChatGPT and DALL-E as separate apps. But in October 2023, OpenAI enabled image generation within ChatGPT for the first time ever.

Just like ChatGPT itself, DALL-E is an extension of the underlying GPT large language model. The latter allows DALL-E to understand the meaning of your prompts regardless of language and convert it into digital artwork. Mind you, this isn’t a novel concept anymore — dozens of AI image generation tools now exist, as we’ll explore in a later section.

Coming back to ChatGPT integration, though, it means you can now type in a description like “A fluffy dog-shaped cloud set against a clear blue sky above a bustling metropolis.” And within seconds, the chatbot will generate four images for you to choose from. Your prompt plays a big role in determining the final image, so you’ll want to specify as much detail as possible.

Can ChatGPT 3.5 or 4 generate images?

Unfortunately, ChatGPT 3.5 or the free tier cannot generate images or AI-generated art. You’ll need an active ChatGPT Plus subscription, which currently costs $20 per month. That may seem steep at first glance but it also unlocks higher-quality responses via GPT-4, plugins, and internet browsing capabilities, among other features.

ChatGPT’s free tier uses the GPT-3.5 language model, which wasn’t created with AI image generation in mind. It’s a fine-tuned version of the GPT-3 model released in 2020, with the newer model simply gaining the ability to handle human-like conversations. Even though OpenAI was working on DALL-E 2 around the time of ChatGPT’s initial release, it remained an independent text-to-image platform.

It wasn’t until late 2023 that OpenAI introduced DALL-E 3 to the world. This time, the company described it as being “built natively on ChatGPT” and that it “lets you use ChatGPT as a brainstorming partner and refiner of your prompts.” Put simply then, ChatGPT improves your image descriptions and passes along a more detailed prompt to DALL-E 3 behind the scenes.

We don’t know how much of this interdependence between DALL-E 3 and ChatGPT depends on the latest GPT-4 language model. Could image generation eventually come to the free version of ChatGPT? Perhaps, but it doesn’t look likely anytime soon. Similar to how GPT-4 has remained exclusive to ChatGPT Plus, image generation could remain locked for the foreseeable future.

How to generate images with ChatGPT

Generating images within ChatGPT takes just a few seconds, here’s how you can get started:

1. Log into ChatGPT via a web browser or the official mobile app.
2. If you haven’t already, sign up for a ChatGPT Plus subscription. DALL-E 3 is limited to ChatGPT Plus users for now, so you cannot use the free version.
3. At the top of your screen, look for a dropdown menu that reads “GPT-3.5” and switch to GPT-4. This is a newer and more capable language model that includes support for DALL-E.
4. That’s it, you can now ask ChatGPT to generate images using natural language. You may want to start your prompt with “Create an image of…” but the chatbot will also generate images in response to questions that benefit from visual representations. For example, if you ask for examples of good looking PowerPoint presentations, it’s likely that ChatGPT will present you with a few images even though you didn’t explicitly request it.

What are some of the best free AI image generators?

While AI image generation is a cool feature, very few people will benefit from it enough to justify a monthly subscription. That’s not a problem, though, as you can use one of the many free AI image generators competing with ChatGPT and DALL-E. Here are your options:

• Microsoft Copilot: Microsoft invested billions of dollars into OpenAI, and relies heavily on the startup’s technology for many of its AI products. That includes its Copilot chatbot that also uses the GPT-4 and DALL-E 3 models under the hood. The best part is that Microsoft lets you generate images for free.
• Meta Imagine: Facebook’s parent company Meta has developed its own image generation model dubbed Imagine. It trained this model on millions of public social media images, which means it can recreate human features exceptionally well.
• Google Bard: Given that Google’s researchers once developed the technology powering OpenAI’s GPT family of language models, it was only a matter of time before the search giant also offered an image generator of its own. And sure enough, Google updated its Bard AI chatbot with the ability to generate images in early 2024 — no payment necessary.

FAQs