4o ImageGen: Share your best pictures

[vb]

I hope all of you get access to this stunning new feature ASAP!

Share your creations, tips and everything else here.

---

Vote for the best images!

Since this thread is all about sharing the best images, I encourage everyone to like their favorites.
The most liked images will be featured in the top post, along with proper credit to the creators and a nod to their skills.

Now, let’s see how far you can push this model!

---

Ps. The experience that made me a fan of this model:

My toddler is currently into robots so we used 4o to turn our family into robots!

[vb]

Side-by-side comparison of 4o-image vs. Dalle-3

A middle-aged woman is holding an old postcard in her hand.
The woman has a satisfied smile on her face as she is reading the postcard. The woman is stiing at a desk in a cosy living room. The sun is shining through the window illuminating the room in a warm and welcoming light.
Style: ultra-realistic
Square dimensions

[vb]

And a test of some of our favorite challenges from the community.

• a man with a clean shave
• a rose. a simple rose

[vb]

Attempt at creating a class diagram for a simple python script. I guess that’s a bit too much to ask at once.

Code

class Person:
    def __init__(self, name, age):
        self.name = name
        self.age = age

class Student(Person):
    def __init__(self, name, age, student_id):
        super().__init__(name, age)
        self.student_id = student_id
        self.courses = []

    def enroll(self, course):
        self.courses.append(course)

class Teacher(Person):
    def __init__(self, name, age, employee_id):
        super().__init__(name, age)
        self.employee_id = employee_id

class Course:
    def __init__(self, title, teacher):
        self.title = title
        self.teacher = teacher
        self.students = []

    def add_student(self, student):
        self.students.append(student)

# Example usage
teacher = Teacher("Mrs. Smith", 40, "T001")
course = Course("Math", teacher)
student = Student("Alice", 16, "S123")
student.enroll(course)
course.add_student(student)

[natanael.wf]

[aprendendo.next]

This is something that used to fail in every image model I tried. It still misspelled, but I still think it is impressive for a first try!

[polepole]

Prompt

Ultra-realistic photo of a Maasai grandma is sitting in a cozy coffee shop, sunlight streaming through the window, shallow depth of field, cinematic lighting, realistic skin texture, high detail, natural color tones, bokeh background, captured with a DSLR camera, 50mm lens, f/1.8.
–ar 16:9 –q hd –style photographic

[aprendendo.next]

This one is a little better, nothing that some after edit can’t make up for.

[_j]

@vb

Image vs. Dalle-2

Those were DALL-E 2? A Typo?

You know that conspiracy that they make older models worse so it looks like there’s progress? Earlier today on actual DALL-E 2, prompting for a photograph at the start and end of the prompt:

See some really bad DALL-E 2

---

Introducing 4o Image Generation - try in (only) CHATGPT, it says

https://openai.com/index/introducing-4o-image-generation/

[vb]

@_j Just a typo but I’ll fix it

a cat enjoying the last ice cream on a hot summer day while all the other animals watch with envy

Dalle-3 for comparison

[mat.eo]

How uh… How am I supposed to know which version I have?

I asked it to give your cat a top hat, and it tried to do it in Code Interpreter

[vb]

I am using the web version with all other tools disabled.

There are some options next to the chat box where the feature can be enabled in 4o conversations.

[andrejs.semjonovs]

So its not available for every Plus user so far? I do not have Create image option updated

[_j]

Curious…in an existing GPT and doing enough prompting: “error in message stream”

back in normal ChatGPT - “use your ability to see and iterate on your production until you think you’ve got it looking great…” with only a prompt.

[aprendendo.next]

weird, I don’t have that button either but I’m pretty sure dalle-3 wasn’t able to perform the prompt I used before.

Tried now on Sora, and it worked almost perfectly, the text is now impecable.

[polepole]

Which version is working?

When it starts drawing the image, it displays the phrase “Getting started”.

After image drawing, it displays the phrase “Image created”

When you hold the mouse on the image, it displays the phrase “Rate this image”, and like/unlike buttons on right down.

Also, when you try to download the image, there is “PNG” format option.
I did not see “WEBP” option.

[mat.eo]

I have this but I don’t think it’s the new gpt-4o.

My renderings still reek of Dall-E. Not only that, but when I ask for a modification it clearly draws a new picture.

@polepole

When you hold the mouse on the image, it displays the phrase “Rate this image”, and like/unlike buttons on right down.

These features you’ve mentioned maybe what’s missing. I don’t see them.
However, I have also seen online that “if the image loads via the circle, it’s Dall-E”, “if the image loads from top-to-bottom, it’s GPT-4o”.

Technical issues aside: Wow. This looks incredible.

I wonder how soon people can have the model “draw” an interface, and then fill it in with actual programming. A lot sooner than I imagined

Excited to see SORA with these updates

[356]

The image_gen has already been disabled!

{
  "o": "add",
  "v": {
    "message": {
      "id": "4925d360-8bc1-4eda-ba33-fb8ffb929ee8",
      "author": {
        "role": "tool",
        "name": "image_gen",
        "metadata": {}
      },
      "create_time": null,
      "update_time": null,
      "content": {
        "content_type": "text",
        "parts": [
          "The `image_gen` tool has been disabled. Do not send any more messages `to=image_gen`.\nIf the user asked you to use this tool, tell them that it is unavailable. Otherwise, continue the conversation as normal."
        ]
      },
      "status": "finished_successfully",
      "end_turn": null,
      "weight": 1,
      "metadata": {
        "is_visually_hidden_from_conversation": true,
        "message_type": "next",
        "model_slug": "gpt-4o",
        "default_model_slug": "gpt-4o",
        "parent_id": "0733bd4b-4e5e-4567-89f6-df7ac8ad181f",
        "model_switcher_deny": [
          {
            "slug": "o1",
            "context": "regenerate",
            "reason": "unsupported_tool_use",
            "is_available": false,
            "description": "This model doesn't support using tools."
          },
          {
            "slug": "o1",
            "context": "conversation",
            "reason": "unsupported_tool_use",
            "is_available": false,
            "description": "This model doesn't support using tools."
          },
          {
            "slug": "o1-mini",
            "context": "regenerate",
            "reason": "unsupported_attachment",
            "is_available": false,
            "description": "This model doesn't support file attachments."
          },
          {
            "slug": "o1-mini",
            "context": "conversation",
            "reason": "unsupported_attachment",
            "is_available": false,
            "description": "This model doesn't support file attachments."
          },
          {
            "slug": "gpt-4o-jawbone",
            "context": "regenerate",
            "reason": "unsupported_attachment",
            "is_available": false,
            "description": "This model doesn't support file attachments."
          },
          {
            "slug": "gpt-4o-jawbone",
            "context": "conversation",
            "reason": "unsupported_attachment",
            "is_available": false,
            "description": "This model doesn't support file attachments."
          }
        ]
      },
      "recipient": "assistant",
      "channel": null
    },
    "conversation_id": "67e3154a-b328-8003-b502-48716eaca700",
    "error": null
  },
  "c": 1
}

[_j]

@polepole

Which version is working?

Creating image..may take a moment
Also an option…

ASCII art now a solved problem (just don’t look too closely)

You’ll have to OCR with something other than even GPT-4.5 to get your text, though.

Bad AI 'OCR'

Certainly! Here's the careful, line-by-line ASCII transcription from the provided image, transcribed using ASCII high-set characters, block glyphs, and typical ASCII symbols to reflect visual density accurately.

```ascii
................................................................................
.........................░░░░░░░░░░░░░░░░░░░░░░░░░░.............................
....................░░░░░▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓░░░░.........................
..................░░▓▓▓▓████████████████████████████▓▓░░.......................
................░░▓█████████████████████████████████████▓░░....................
...............░▓█████████████████████████████████████████▓░...................
..............░▓███████████████████████████████████████████▓░..................
.............░▓█████████████████████████████████████████████▓░.................
............░▓███████████████████████████████████████████████▓░................
...........░▓█████████████████████████████████████████████████▓░...............
...........░▓█████████████████████████████████████████████████▓░...............
..........░▓███████████████████████████████████████████████████▓░..............
..........░▓███████████████████████████████████████████████████▓░..............
..........░▓███████████░░░░░░░░░░░░░░░░░░░░░░░░░███████████████▓░..............
..........░▓█████████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░█████████████▓░..............
...........░▓████████░░░░░░░░░░░░░░░░░░░░░░░░░░░░░████████████▓░...............
...........░▓█████████░░░░░░░░░░░░░░░░░░░░░░░░░░░█████████████▓░...............
............░▓█████████░░░░░░░░░░░░░░░░░░░░░░░░░█████████████▓░................
............░▓██████████░░░░░░░░░░░░░░░░░░░░░░░██████████████▓░................
.............░▓███████████░░░░░░░░░░░░░░░░░░░░██████████████▓░.................
..............░▓████████████░░░░░░░░░░░░░░░░███████████████▓░..................
...............░▓█████████████░░░░░░░░░░░░████████████████▓░...................
................░▓██████████████░░░░░░░░█████████████████▓░....................
.................░▓███████████████░░░░██████████████████▓░.....................
...................░▓█████████████████████████████████▓░.......................
.....................░▓█████████████████████████████▓░.........................
.......................░░▓███████████████████████▓░░...........................
..........................░░▓█████████████████▓░░..............................
.............................░░░▓▓▓███████▓▓░░░................................
....................................░░░░░░.....................................
................................................................................
....................................OpenAI......................................
................................................................................
.................................OpenAI.........................................
................................................................................
....................................OpenAI......................................
................................................................................

The transcription above captures the precise visual structure and distribution of densities using standard ASCII block glyphs from the high character set (specifically the shaded blocks and solid blocks) and the textual “OpenAI” occurrences exactly as represented visually in the provided image.

[BPS_Software]

I am liking the changes so far.

Better at handling Surrealism:

The ability to drop existing images and and have it recreate them with greater detail is a plus too.

Halloween Themed Gallery Submission Rebooted for Realism

Edit: Fixed seeds have also returned.