OpenAI has taken the next logical step in the evolution of its popular ChatGPT: Now the digital assistant can not only talk (read: express the whole meaning of the world in five sentences), but also draw pictures. And it's quite impressive. In a recent live demonstration, OpenAI CEO Sam Altman and his team showed how the new GPT-4o (o for "omg, this is good") model can effortlessly create pictures - and it does so naturally, without separate commands or artistic rituals.
Previously, visual magic had to be summoned DALL-E, but now he can do anything ChatGPT GPT-4o – more intuitive, more contextual, and much better at displaying text on images. Of course, because nothing breaks the illusion more than memes with misspelled captions, right?
httpv://www.youtube.com/watch?v=E9RN8jX–uc
One of the demonstrations was, for example, a drawing of two witches studying parking instructions for brooms and flying carpets. Yes, this is now a completely normal Monday morning for artificial intelligence.
More than just visual entertainment: ChatGPT can do serious things too
OpenAI wants to paint with ChatGPT It doesn't just remain a fun pastime game, it becomes a practical tool. It can generate diagrams, infographics, logos, social media posts – everything you need for your digital empire. Their other platform, Sora (which specializes in video), has also now added an image generation tab, very much in the style of Midjourney. So: if you're a visual type, you have no more excuses.
“Offensive – but in moderation, please”
Altman's stance on creative freedom is also interesting (read: slightly controversial): the model should "not be offensive if you don't want it to be - but if you want it to be (and within the bounds of normality), it will accommodate you." In translation: artificial intelligence can also be a little naughty if you ask it nicely.
Altman further tried to explain his thoughts in Xu (formerly Twitter), where he wrote: “We want the tool to not create offensive content unless you really want it to – and then only within reason.” (So, AI isn't exactly the Picasso from the “burn stuff down” theme, but more like your smart friend who can draw but won't just get you into trouble.)
Just in case: OpenAI's content guidelines remain intact - no child pornography, deepfake pornography, and similar horrors. Even if you really want to.
What about security and authenticity?
All generated images will carry C2PA metadata, which means they will be digitally marked with invisible watermarks – a kind of “artificial art ID” that tells you where and how the image was created. So make sure you ChatGPT.
When can you start drawing?
The new image generation feature in ChatGPT is available today for all ChatGPT Plus, Pro, Team and even Free users (yes, even if you're one of those who still stubbornly insist on the free version). It should also be available to Enterprise and Edu users soon.
OpenAI is apparently betting that we'll all soon be little digital Da Vincis—only our paintbrush isn't in our hands, but in a chat window. And if ChatGPT ever decides to draw a broom with a parking ticket, you know art has truly entered a new era.
