OpenAI previewed a new version of their AI image generator DALL·E (v3) this week, and they also made a change in how users access it. When comparing DALL·E to Midjourney or Stable Diffusion, there are 3 main things that make it different (at least for now):
- DALL·E can render text, which all the other popular image generators struggle with.
- DALL·E doesn’t require elaborate prompt engineering. All it needs is simple descriptions of what you want.
- DALL·E will be integrated within ChatGPT right inside the browser or app, a welcome change to using Midjourney in Discord.
Number 2 is the big one… users will not be required to be a “prompt engineer” for much longer. If you've used or even tried Midjourney via Discord you know what I mean. It's a mess.
The other noteworthy advancement here is that ChatGPT is now multimodal—it can understand and generate both text and images. It will be interesting to see how the competition responds to these advancements with features of their own.
Artists and creators can also opt out their content from future training models, and DALL·E steers away from making ethically-controversial images "in the style of" more well known artists by default, although I'm sure the internet will figure out a way around this.
DALL·E 3 is designed to decline requests that ask for an image in the style of a living artist. Creators can now also opt their images out from training of our future image generation models.
Access to these features will come this Fall, and it sounds like DALL·E will be available inside Bing Chat for free courtesy of Microsoft’s partnership with OpenAI.
Read more in this post on their site.
Also worth noting: the cute little hedgehog feature video is an interesting strategic approach when a company is trying to disarm the public perception of the potential threat that AI might literally be taking over the world right now. Just look how cute it is!