ChatGPT Starts to Break the Bounds of Chat
Rachel Metz
What started as a DALL-E obsession for me way back in the day eventually morphed into a Midjourney addiction. But in recent months, Gemini became my go-to AI image generator. Why? 'Nano Banana' of course. Just exceptional work out of Google – including going with the correct stupid name for maximum virality.1 OpenAI, deep in the throes of a "Code Red", wants their crown back:
OpenAI is rolling out an update to ChatGPT that’s intended to generate images better and faster, the latest move by the artificial intelligence developer to bolster its flagship chatbot amid heated competition from Alphabet Inc.’s Google.
The new version of ChatGPT Images, announced on Thursday, is designed to make and edit images more precisely, as well as to spit out pictures as much as four times faster than its previous AI image-generation model. The company is also creating a new section within ChatGPT’s mobile app and website meant for people to make images, rather than just doing so in an interaction with the chatbot.
I'm happy to report that it's good. Very good, even. I'm only a day or so into using it, but I'm finding myself preferring it over Nano Banana right now in my early side-by-side tests. Most importantly, it's fast. Still not quite as fast as Gemini in my usage, but the biggest issue that ChatGPT had previously with images was just how sloooow it was to generate them. It was tedious. This new version is not.
But the bigger change may actually be the UI tweak to ChatGPT itself alongside this new image model. I always found it a little odd that image creation was a part of the chat interface. On one hand, that's good for simplicity. But we're to the point now where ChatGPT is a pretty robust product, with multiple features and functions. While media you created was sorted into its own area, now image creation has a native home there too. You can still do it all from the main text prompt, but for most users, this area will probably make more sense.
Fidji Simo's post on the matter explains the company's rationale:
Over the past few months, I’ve talked about how ChatGPT is evolving from a reactive, text-based product into something more intuitive and connected to any of the tasks you want to accomplish. The shift from text to multimedia and dynamic UI is an important part of that transformation, and I’m excited about the progress we’re making.
|Many people’s first experience with ChatGPT involves turning a text prompt into a picture. It’s a magical way to see what this technology can do, but the chat interface wasn’t originally designed for this. Creating and editing images is a different kind of task and deserves a space built for visuals. Today we launched a new image gen model and a dedicated entrypoint in ChatGPT for images that works more like a creative studio. The new image viewing and editing screens make it easier to create images that match your vision or get inspiration from trending prompts and preset filters.
And yes, this does seem to be just the start of reworking the ChatGPT UI to better accommodate the job being done, rather than just, you know, chat.
Again, there are risks to that move away from simplicity, but it also feels like the right time to evolve. Especially when Gemini's own product has evolved to look... a lot like ChatGPT. One of OpenAI's key strength has been their ability to productize AI, and Google, it seems to me, is starting to catch up. So you gotta keep moving.2
1 Sort of ironic given OpenAI's use of fruit nicknames-that-should-have-been-product-names as well... ↩
2 Especially if one end-goal is to create new kinds of devices that lives far beyond a text prompt... ↩