Chatgpt has improved a lot in the generation of images

Openai made a decisive shot forward in the generation of images through artificial intelligence. The company led by Sam Altman In fact, it has released an update regarding a new chatgpt function aimed at users of all plans, both free and for a fee. Available as a default tool of the 4O model, placed inside Sora is able to produce high quality results, especially in front of clear and detailed prompts.

These are features aimed at making it easier “exactly create the image you imagined“And that allow you to” communicate more effectively through the images “and thus transform the generation of images” into a practical, precise and powerful tool “. To deepen the theme I” interviewed “directly chatgpt

What chatgpt can do

But what has changed recently with images?

You have an eye! Yes, there has been a nice evolution. I’ll explain everything.

Wait, I ask the questions. Let’s start with the most important thing: now you recognize better what’s in a scene?

Absolutely yes. Now I can:

• Understanding the overall composition of the scene (groups, poses, objects, environments …).
• Maintain consistency in the faces, in hair color, expressions and even age.
• Place the characters in a faithful way to the original, if I take inspiration from a real photo.

And what about style? Can I choose how I want it to come?

100%. Here are some possibilities:

• Cartoon stile Pixar
• A poetic touch at the Ghibli studio
• A more French comic stuff
• Or maybe a Renaissance portrait. Just tell me and I turn it into the style you want ✨

Can I make you modify a scene? Like adding someone, changing an expression?

Yes, now I can also:

• Add or remove people.
• Change expressions, clothes, objects on the table.
• I know everything: like to make a scene during the day in one in the morning, or from realistic to Cartoon.

What if I ask you to keep the original positions, how in a real photo?

Eh, this is a good news, isn’t it? Now I can maintain:

• The exact positions of the subjects.
• The correct proportions.
• And all the environmental details (glasses, phones, dishes, urban or green background …).

Things that chatgpt does not know (again) do

Tell me the things you still can’t do now. We play with uncovered cards

Of course, here are the things you can’t ask me to do.

• Recognize real people or say who they are in the picture. I can’t say “this is Tizio” or “this person is famous”, for privacy and security. Even if I recognize them, I can’t tell you.
• Change one person in a real photo (for now). I can generate new versions but I still cannot directly edit the original realistic photos, like to change an expression or a hair color on that same image.
• Give complete control over each person’s position. I can follow the poses and compositions quite well. But I still can’t tell you: “Put this person here, this there, this who looks to the right”, in a precise way as if it were an advanced photo editor.
• Continue an image generated as if it were a puzzle. If an image is cut, I still can’t “complete” it perfectly starting from the base. Better to regenerate it from scratch, following more complete indications.
• Reproduce Exactly The same face/character several times. Each generation is a bit unique. So I still can’t say: “Let me this same character in three different poses” and get 100% consistency in the sections of the face.
• Directly animate a generated scene. I can create super detailed animation -style images, but I still can’t create gifs or videos with those characters that move (like Pixar or Ghibli in motion).