Released general. THE OpenAI requires no more that those interested in playing with the Dall-E 2 sign up for a waiting list for this. The announcement, made this week, means you only need to log in to the AI page and it will create images based on what you type.
Dall-E 2 is a machine learning-based tool that generates images from the words the user enters in a field. There are others with the same proposal, like Midjourney and Stable Diffusion (which can even be used to compress images, look that). All are so easy to use that they have become a hobby on the web.
could. Many images generated from the text instructions are so impressive that they are considered works of art. Others are so bizarre that they end up being fun.
The Dall-E Waitlist
Responsible for the project, OpenAI presented the first version of Dall-E in early 2021. The results were so impressive that they pleased both experts in artificial intelligence and the lay public.
But the growing interest in the tool could bring problems for OpenAI. Depending on the combination of words, Dall-E could generate images with a negative context, which would serve to disinformation, prejudice and embarrass people, for example.
This is the main reason why OpenAI created a waitlist for Dall-E. While keeping access to the service in check, the organization tried to implement filters to prevent the system from creating problematic images.
Such filters prevent — or try to prevent — the Dall-E from generating images with a sexual, violent or racist content, for example.
There is criticism for the way OpenAI handled the problem. To The Vergethe organization confirmed that, in addition to activating filters, it can add invisible words to users’ instructions to avoid potentially harmful results.
For example, Dall-E can insert “Asian woman” into a word set that involves people but does not specify gender or ethnicity. It is a way to prevent only whites or men from appearing in the results. However, this approach has been criticized for often leaving the images far from the expected result.
Anyway, the OpenAI considers that its security mechanisms are already competent enough for Dall-E 2 to be released to everyone. The organization also tests an API to allow the technology to be used in third-party applications or plugins.
If you want to test the latest version of the technology, you just need to create an OpenAI account and log in to Dall-E 2 website. By accessing the service, you receive 50 credits. After that, 15 more credits are provided every month.
Each credit allows you to generate a round with four images, edit an already generated image (inpainting) or extend an already created image (outpainting) for new results. If the free credits are not enough, you can buy more.
Within the service, just type words in the main field of the site and wait for the result. To generate the images just above, I typed some objects that were on my desk: bottle (bottle), keyboard (keyboard), book (book) and printer (printer).
Cool breeze, huh?
https://tecnoblog.net/noticias/2022/09/29/agora-todo-mundo-pode-usar-o-dall-e-para-gerar-imagens-a-partir-de-texto/