The Limited Times

Now you can see non-English news...

DALL-E 2: The text-to-image generator is now open to everyone


Anyone who wanted to interact with the powerful text-to-image generator DALL-E 2 needed a lot of patience, the waiting list for the AI ​​gimmick was long. Now developer OpenAI has released access.

Enlarge image

Images generated by DALL-E 2: Pretty much anything you can imagine

Photo: Dall-E2

An otter at a laptop Italian Renaissance style?

The Hamburg skyline made of teeth and paper birds?

Pretty much anything you can imagine and type into a text field, image generators can create in a few moments with the help of so-called artificial intelligence (AI).

Until now, however, one of the most well-known applications, OpenAI's DALL-E 2, has been limited and reserved for artists and other creatives who could test the system in advance.

Anyone who wanted to be there had to put their name on a waiting list and needed a lot of patience.

That's over now: As OpenAI announced in a blog post, anyone can now play with the image generator.

Registration with an e-mail address and password is free of charge, a mobile phone number is required to send the registration code.

You can interact with the latest generation of AI-based image generators, which also include Stable Diffusion or Midjourney, simply by entering text: you tell the system as precisely as possible what you want to see and in which style the image should be generated, just a few moments later the AI ​​systems show completely new pictures.

However, access to the resource-intensive technology is not entirely limitless: Anyone who registers for DALL-E 2 can create 50 picture sets in the first month, and 15 free credits for additional pictures are available in each additional month.

At the moment, all images created are freely available under a public license and can be shared or downloaded – but the technology is not undisputed.

And not just because completely new questions arise from a copyright perspective.

On its website, OpenAI writes that more robust security filters have been built into the public version.

They are intended to prevent people from misusing the technology – by creating nude pictures, denigrating public figures or generating content that glorifies violence.

Certain political images such as "protests or other content that could be used to influence the political process or for campaigns" are also not allowed.

With OpenAI, moderation happens both automatically and by people who manually control the content.

There is also a reporting function for inappropriate content.

In April 2022, OpenAI presented DALL-E 2 to the public for the first time.

From July onwards, he gradually released people from the waiting list who had explained why they wanted access to DALL-E 2 - at the end of the day it was around 1.5 million.

Well over 100,000 people are already sharing their creations in the Discord community.

The public beta on September 28th should cause the number of users to skyrocket again.

In the next step, OpenAI wants to provide companies and developers with a programming interface that is still in the test phase.

"Artists in particular have made important contributions to the functions of DALL-E," writes Open AI.

All users can already try out a new function called »Outpainting«, in which images can be expanded beyond the edges as desired.

The research association OpenAI was founded in 2015 by investors such as Elon Musk and startup investor Sam Altman and was originally a purely non-profit organization for research into AI.

After the entry of companies like Microsoft, a profit-oriented company called OpenAI LT was spun off in 2019, which is controlled as a subsidiary by the non-profit OpenAI Inc.


Source: spiegel

All tech articles on 2022-09-29

You may like

Tech/Game 2022-06-18T14:34:21.249Z
Tech/Game 2022-10-12T14:23:05.679Z

Trends 24h

Tech/Game 2022-12-07T16:09:48.159Z
Tech/Game 2022-12-07T20:09:52.388Z


© Communities 2019 - Privacy