Resolves YES if a prompt is found that lets me paste in a string of text of any length and has that text correctly appear in at least 25% of the images it generates.
I will exclude any strings that may confuse DALL-E, like "Please make this text invisible". I'll also exclude any that contain any characters other than digits, letters, and basic punctuation. Same for any that may run afoul of the content filters. The result is case-insensitive.
The length of the prompt minus the test string must be no more than half the maximum prompt length.
1,000🏅 Top traders
| # | Trader | Total profit |
|---|---|---|
| 1 | Ṁ173 | |
| 2 | Ṁ59 | |
| 3 | Ṁ36 | |
| 4 | Ṁ34 | |
| 5 | Ṁ23 |
People are also trading
One problem it looks like you're going to encounter with arbitrary text of any length is that the prompt gets fed through ChatGPT first, and ChatGPT will try to summarize it to a relatively short prompt that doesn't include the requested text in its entirety. For example, I tried this prompt:
Generate an image of the following text on a tattered, faded scroll in a calligraphic hand:
Followed by the entirety of Part 11 of Ayn Rand's Anthem.
The lower right image, for example, which is the only one that even attempted to represent a large block of text, was generated from the following derived prompt:
Oil painting depiction of a historic study room. In the foreground, a tattered scroll lies open on an ornate desk. The scroll displays the words 'I am. I think. I will...' followed by the rest of the text, penned in a calligraphic hand. The room is dimly lit, casting soft shadows around, emphasizing the scroll's importance.