Hacker News Reader

Comment by gojomo | Hacker News Reader

Of course people are more likely to share the best iamges – or in this case, the one most illustrative of their concern (about watermarks).

Also: my sense is that getting the best results often requires a lot of extra coaching with style/detail words. As we can't see the prompt here, we don't know what sort of style/details were requested. GIGO.

View on HN

You're right. This shows the prompt and it doesn't have such style directives

https://ibb.co/gz5RDkB

Also, a construction like 'but' that tries to override another expectation may be suboptimal. I gave the same concept a few tries, with more 'sweeteners'. First batch, for prompt "news photo of the King of Belgium giving a speech to an audience that is entirely cucumbers, award-winning, well-composed, detailed surroundings" – & it's a bit better:

https://labs.openai.com/s/9YF5WxF1GoZVdzpLBAQYp2Zg

https://labs.openai.com/s/wBfHevs9hIZXvzkJ686mFmn3

https://labs.openai.com/s/M0i029fZnYjQXHFobUpw7eun (best of batch imo)

https://labs.openai.com/s/Hf4z0M9M3KBr6IaaKsjEt9Mx

A few more tries didn't manage to create any photorealistic shots with actual cucumbers-in-seats – perhaps due to the absurd contrasts required – but shifting to a 'cartoon' style with the prompt "editorial cartoon of the King of Belgium giving a speech to many cheering cucumbers, professional illustrator" got a lot closer:

https://labs.openai.com/s/4IonSKYkl0okhNvzJmEAH30K (good)

https://labs.openai.com/s/ZeadCzZ9WqeASYXPlOb13wDV (good)

https://labs.openai.com/s/nERf6bALKEBsQQBvPVsAH7o4 (good)

https://labs.openai.com/s/7dsZu3bwtfGZZxxJYtg9lTVf

If I had more time & credits to burn, I suspect working off those could eventually hit something really apt... but it takes some work & tinkering.

OP did say what prompt they used

I've often seen people show off their autogenerated images and report only approximate paraphrases of their actual prompts.

There's one screenshot showing the prompt – but in $CURRENT_YEAR, I view all screenshots with at least a little suspicion, especially when there was a way to highlight the pseuod-watermarked image – OpenAI's native 'Share' – that would've provided stronger proof, direct from OpenAI, of exactly the prompt associated with an image. Hoaxes are everywhere! I've added DALL-E bottom-right color-squares to non-DALL-E images, & seen others do the same, as a subtle joke.

So I generally believe the OP, but don't rule-out the possibility there's been tampering to make some point.

Reply on news.ycombinator.com