News for hackers. Stuff that matters.

A comment on Ask HN: DALL-E was trained on watermarked stock images?. Parent comment

Re:Ask HN: DALL-E was trained on watermarked stock images? by lioeters on Thursday August 25, 2022 @01:17AM (#32589801)
> you have to think of the underlying labeled text-to-image sets as paint colors to mix, and prepare a palette accordingly.

Very insightful tip on how to harness the "creativity" of Dall-E and the like.

I see how the phrase "king of belgium" was too vague for Dall-E, so it didn't produce anything recognizable - but changing the words into known details, like "banker" and "salt and pepper hair", worked effectively to generate concrete imagery.

Hilarious results. :)

[ Parent | on HN ]
Re:Ask HN: DALL-E was trained on watermarked stock images? by ehsankia on Thursday August 25, 2022 @03:43PM (#32599004)
It's not that it's "vague", they intentionally throw off when you try to generate a photo of a named person. It's an intentional protection they put in. If you just do "king" it'll likely do fine, but if it's referring to a specific person it won't.
[ Parent | on HN ]
Re:Ask HN: DALL-E was trained on watermarked stock images? by lioeters on Thursday August 25, 2022 @04:04PM (#32599279)
Ah I see what you mean - "king of belgium" is a real person, so they put in some safe guards in DALL-E to prevent recognizable images for such queries. Makes sense.
[ Parent | on HN ]