This article introduces a new generation of text-to-image models presented by OpenAI, and the results are stunning

OpenAI presents a new generation of text-to-image models, and the effect is stunning

This article was original written by Jin Tian, you are welcomed to repost, first publish at blog.tsuai.cn. Please keep this copyright info, thanks, any question could be asked via wechat: jintianiloveu

Let’s start with a few pictures:

In simple terms, the task is to type a text and have the AI draw it, such as:

a pentagonnal greenn clock. a green clock in the shape of a pentagon.

Draw a green pentagonal green clock, drawn by AI:

Look at the green, how green it is, and the luster of the clock, without any PS traces!

For example, let’s give a topic like this:

a snail made of harp. a snail with the texture of a harp.

We want to keep a snail that looks like a harp, yeah, you heard that right, I want this snail to look like a harp.

Look at what AI has given you. This level, can the Central Academy of Fine Arts not admit you? A snail like a harp!

What struck me the most was that this model, oh no, this AI, it had this genius hand that could draw, and it was full of imagination! For example, let’s ask him to draw this:

an illustration of a baby daikon radish in tutu walking a dog

Do you know what that means? Draw a picture of a turnip child walking a dog in a short skirt.

Radish child?? How would you draw it? It takes imagination!

Look at what AI has drawn for you! Do you have the radish kid yet? The short skirt, walking the dog, all the ingredients, and look at this radish kid, it really looks like a radish kid! The dog also looks like a dog. The AI is sure there is a real person behind it.

But, yes, that’s what OpenAI is doing these days: Dall-E. All the drawings are made from this model.

You can go to OpenAI’s official website: openai.com/blog/dall-e… This is literally the strongest AI model I know of since GPT-3.