Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I'm over 1000 credits into Dalle so far (I know, I know..) and you're on the money. You can go a lot further than impressionism, though. Specifying the names of famous illustrators, photographers or artists. Specifying the media used. Lens types. Colorways. Film types. Lighting. The right combinations can yield some incredibly realistic looking things, even faces, and then for the rest of it, there's Photoshop, Gigapixel, and other tools to patch things up. (I've had more luck creating 'elements' with Dalle and then montaging them the old fashioned way.)

The images used in the blog linked by OP are okay but stylistically all over the place. OP acknowledges how difficult good prompts are to write. Beyond that, though, you still need to think like an art director and establish a way to set a common style to avoid jarring the readers, and Dalle alone can't do that.



Dall-E is very “first generation” in its design and interactivity with the users.

It’s just a matter of time until setting a “consistent art style” becomes a feature of these things.

Similarly, asking the AI to produce multiple views/poses of the same thing will likely become a common feature.


> Specifying the names of famous illustrators, photographers or artists. Specifying the media used. Lens types. Colorways. Film types. Lighting.

Are training sets prepared with systematic variations in individual axes, as an alternative/addition to tagging each of millions of training images on these axes?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: