OpenAI done did it again, y’all! They dropped some new image-making magic for developers to play with in ChatGPT. This fancy image generator has been making waves since late March, spitting out Ghibli-style pics and AI action figures like nobody’s business. It’s been a blessing and a curse for OpenAI, bringing in a bunch of new users while also putting a strain on their servers. Over 130 million ChatGPT peeps made over 700 million images in just one week – talk about a creative frenzy!
In the guts of OpenAI’s API lies an AI model called “gpt-image-1” that’s like a chameleon of image-making. It can whip up pics in different styles, follow rules, use knowledge from the world, and even add text. Developers can churn out multiple images at once with this bad boy and tweak the quality to control how fast things get done.
Now, here’s the kicker: gpt-image-1 plays by the same rules as ChatGPT when it comes to safety. It won’t spit out anything that goes against OpenAI’s policies. Developers can choose how strict they want the moderation to be, from “auto” to “low” filtering. Low filtering means fewer restrictions on what the model can cook up, according to OpenAI’s notes to TechCrunch.
And get this – all images made with gpt-image-1 come stamped with C2PA metadata, so folks know they’re AI-made. The price tag for all this image-making goodness? Five bucks for every million text bits, ten bucks for every million image bits, and forty bucks for every million image bits spat out. That’s about 2 cents, 7 cents, and 19 cents per image for low, medium, and high quality snaps, respectively.
OpenAI says big shots like Adobe, Airtable, and Instacart are already messing around with gpt-image-1. Figma Design even lets folks create and tweak images using this tech, while Instacart is trying it out for recipes and shopping lists. Looks like OpenAI’s new toy is making some noise in the tech world!