What the heck is going on with Dalle-2 and the API?

As someone who spent over a year creating an web app that lets people build a virtual world, using an in-world chat box that generates images via calls to the Dalle-2 API, I am perplexed and really disappointed in the state of Dalle-2 and the Dalle-2 API right now.

OpenAI comes out with Dalle which arguably launched the whole AI generated image craze, creates an API for it, and then goes silent on it as the rest of the world picks up the ball and runs with it.

Worse, they tease us with a few images on Twitter/X that are at MidJourney level quality, hinting that a vastly more powerful version of the engine and API are coming soon.

And then… Nothing <crickets>.

And this has been for months! Can someone please tell me if my fear that OpenAI has dropped real support for upgrading Dalle-2, because they are focusing on ChatGPT, is real or not? I’ll be heartbroken if it is because my code now, after many months of efforts, works perfectly now. As it is, I’m already making serious plans to swap out the Dalle-2 API for Leonardo AI’s API and the same goes for MidJourney when they release their API because of OpenAI’s silence on this. Who walks away from the hottest AI software craze ever that they started?

If OpenAI didn’t have the backing of Microsoft, one of the richest and most powerful software companies in the world, and instead they were a small underfunded software company, then I might have seen this coming.

But to say that this is bewildering is putting it mildly! Does anyone have any information on what is going on behind the scenes at OpenAI in regards to Dalle-2 and the Dalle-2 API?

NOTE: I did see this other similar thread on the topic, but there still are no answers from OpenAI so I felt this issue should be raised again to emphasize the frustration of the developer community over this:

1 Like
  1. The company can, and does, work on multiple things simultaneously.
  2. The market for generative text is many orders of magnitude larger than the market for generative images.
  3. I’m unclear what, exactly, you’re expecting from them in terms of a product that is less than a year-old.

API imagery costs 1/4 the credits of labs.openai.com, and doesn’t state “dall-e” or version anywhere in documentation.

So the question is more like: when will labs’ Dall-E (or Bing AI image quality) be an API product?

1 Like

From what I know, labs.openai.com is still using old DALLE2 version.

Earlier this summer, a few thousand people were invited to try DALLE2.exp (experimental) on labs but not via API… They closed that down the end of July and promised that the team is “heads down” and working on final version to release.

The high-quality you saw was likely DALLE2.exp…

It’s been about 6 weeks since they closed the trial period for the experimental model after taking all our feedback, so hopefully we’ll have more news soon.