ChatGPT Now Offers Limited Free AI Image Generation
Toggle Dark Mode
While OpenAI’s ChatGPT bot has always been great fun for having text-based conversations and generating streams of written consciousness, many of its more powerful features have been locked behind a paywall, requiring customers to subscribe if they want to do more.
For example, generating AI images requires at least a ChatGPT Plus subscription, which starts at $20 per month. While that may not be too bad for someone who relies on AI for business, it’s pretty steep for those who just want to play around with generative AI.
The good news is that OpenAI has realized this, and it’s unlocking the ability for those on its zero-cost free plan to make two image generation requests per day without subscribing or paying anything extra. You’ll still have to sign up for an account to use it, but you won’t have to pay.
ChatGPT’s image generation uses the same Dall-E 3 model that rolled out to its paid subscribers last year, so at least free users won’t be hamstrung by working with an older or less-capable model. However, it’s important to note that the daily limit applies to requests and not the actual images.
In other words, you’ll want to be as specific as possible on your first try since if you don’t like what you see, you’ll need to use up your second (and only remaining) request to refine it further. If you still don’t get the results you want after that? You’ll need to pay up or wait until the next day to get two more free requests.
If you think that makes OpenAI’s new policy sound like more of a gateway drug to try and get folks to sign up for a paid tier, you’re probably not wrong. It’s just enough to see what Dall-E 3 is capable of but not really enough to get many useful results unless you’re a zen master at AI prompting.
ChatGPT is also unique in restricting its AI image generation features to paying customers since Microsoft’s Copilot and Google’s Gemini both offer these capabilities on their free tiers, and Microsoft is using the same GPT-4o model at ChatGPT. Like OpenAI, both Microsoft and Google provide better results and more complex AI models to paying subscribers, but the basics are free across the board.
How Does This Affect Apple Intelligence?
Although Apple has partnered with OpenAI to include ChatGPT in Apple Intelligence later this year, that’s only to fill in some of the blanks that Apple’s own AI large language models (LLMs) won’t be up to handling quite yet.
Fortunately, image generation isn’t one of those. While it has yet to appear in the iOS/iPadOS 18.1 and macOS Sequoia 15.1 betas, Apple plans to bring its own image generation tools to the iPhone, iPad, and Mac, which will rely on Apple Intelligence to perform image generation.
These will be handled on-device as much as possible. However, the power required to handle generative AI for graphics makes it likely Apple’s secure Private Cloud Compute infrastructure will also get involved (which may also be why we haven’t seen these features yet — Apple still has to get the back-end ready).
The role of ChatGPT in Apple Intelligence is much narrower than you may think. The separation between Apple’s own AI models and where ChatGPT comes in can best be summarized as the difference between “personal” and “world” knowledge.
Apple Intelligence will be involved in nearly everything that revolves around your personal life. If you ask Siri what time a flight lands, if you can make it to a concert on time, or even for movie and book recommendations from your friends (based on messages and emails), that’s all done using Apple Intelligence. ChatGPT doesn’t come into it.
However, if you want to know how to make a nice pasta dish from the ingredients in your cupboard or you’re looking for home decor suggestions, Siri will have no idea how to help you, as it doesn’t watch The Food Network or read Better Homes & Gardens (at least not yet).
Instead, Siri will offer to send your request over to ChatGPT, which is more capable of providing constructive answers to such questions. You’ll always be told that Siri wants to do this and asked if you wish to proceed — and you always reserve the option to say no. In fact, based on early reports, Apple doesn’t even plan to provide an “always allow” setting for this — you’ll be prompted each and every time Siri thinks ChatGPT should get involved in the process. Plus, you won’t need to have a ChatGPT account to use this, so any requests that do get sent over are entirely anonymous; OpenAI has no way of associating them to you, nor is it able to maintain a persistent identity to tie multiple requests together.
Based on what we’ve seen so far, though, Apple Intelligence won’t hand any image generation tasks over to OpenAI, even if you’re a ChatGPT Plus subscriber. That’s all done with Apple’s LLMs and image generators, so ChatGPT doesn’t need to get involved. Of course, ChatGPT fans can always install the ChatGPT app and generate images with that (as with any other online tool), but Siri’s willingness to engage ChatGPT will have quite a few guardrails around it.
