OpenAI brings image generation to all free ChatGPT users

If your feed felt like a Studio Ghibli storyboard over the past week, you were not imagining things.

OpenAI's latest image generation feature went wildly viral when it rolled out on March 25, prompting CEO Sam Altman to post that their GPUs were literally "melting." The demand was so massive that ChatGPT briefly added one million users in a single hour, forcing the company to delay access for non-paying users.

Altman confirmed on X that the tool is officially available to everyone on the free tier.

Free users get three image generations per day. If you need more, you still have to pay for the $20 monthly Plus plan or the $200 Pro tier. While the cap is tight, this isn't the old DALL-E 3. GPT-4o's native generator delivers noticeably better text rendering, tighter spatial layout, and sharper image editing capabilities right inside the chat window.

The guardrails are coming down

While the Ghibli memes stole the spotlight, OpenAI quietly pulled off an even bigger shift behind the scenes: it relaxed ChatGPT's strict content moderation rules.

Previously, prompt triggers around sensitive topics or public figures met an immediate, blanket refusal. Now, OpenAI says it is ditching outright bans in favor of a targeted model aimed at preventing "real-world harm," according to Joanne Jang, the company's model behavior lead.

ChatGPT can now generate and modify images of public figures like Donald Trump and Elon Musk. Jang noted that OpenAI does not want to act as an "arbiter of status," and instead is offering an opt-out process for figures who don't want to be depicted.

Users can also generate historically sensitive imagery, including swastikas, provided the prompt is framed in an educational or neutral context and does not promote extremism. Prompts adjusting physical features—such as making someone look heavier or adjusting ethnic traits—are now being processed rather than rejected.

To be clear, it is not a free-for-all. OpenAI beefed up protections around imagery involving minors compared to DALL-E 3. And while broad studio styles like Pixar or Studio Ghibli are allowed, mimicking the style of living individual artists remains blocked.

Why it matters

OpenAI just handed its 500 million weekly users a massive creative toy, and capping it at three images a day is a calculated masterstroke.

It gives mainstream users just enough taste of frontier image generation to fuel viral social media trends without completely melting the company's server infrastructure. For casual users, you do not need complex prompting software or a Discord server anymore—you just open an app you already have and generate an avatar in seconds.

Strategically, peeling back safeguards on public figures and sensitive topics is a high-wire act. On one hand, it addresses long-standing complaints about Silicon Valley censorship, especially as competitors like Google previously faced PR disasters over over-sanitized outputs. On the other hand, putting public figure generation into hundreds of millions of pockets right now invites massive deepfake risks. OpenAI is betting that precision filtering can outrun political and legal backlash—we will see if that gamble holds.