Take an image and try to recreate it by writing a poem as the prompt

I started with this image, which was a random choice I'd recently created:

Instead of making something locally using Stable Diffusion as I usually do, I figured this challenge would probably go best with Dall-E 3, due to its advanced prompt following skills (well, so long as you don't offend its tender sensibilities anyway).

Here was my first go, which was really an accident as I hit enter before I really got anywhere with it 😀

But here is the real deal, which I was quite pleased with. Check out the poem below the image.

And the poetic prompt:

On one fine misty spring morning, I took a walkabout
Still in my silvered battle mech armor, atop my black firesuit
A silly sight my slim figure made, looking into the distance
    with a carpet bag over my shoulder!

The trail ends at the grassy shore of a curving brook, edged with rocks big and small

A welcoming sight in the distance! This ancient Chinese traveler
may soon find her rest for the eve

I am met with two rustic buildings, perhaps a tavern and a farmhouse?

The eaves under the upturned roof are beckoning to me

The porch seems to have a chair with my name carved on it

The bushes near, the mountains afar -- I am home

I'm no master of meter, but this did the trick for me.

"Put your face in any style" is the claim made by the authors of the paper titled: InstantID : Zero-shot Identity-Preserving Generation in Seconds. You don't have to know what this means, as the team spent a lot of time providing examples and comparisons to previous techniques. This stuff moves so fast -- I was evaluating one of them the day before writing this article, which had also been recently introduced! 

Our model supports identity-preserving generation in high fidelity with only single reference image in any style.

We can achieve competitive results as LoRAs without any training.

Expect to see even more "face cloning" or "selfie styling" apps and services and features over the next few months. This stuff was already possible for most of a year, but you really needed many photos, careful preparation, and a lot of expensive GPU time to do it well. Not anymore. I have this running on my PC now, takes 30 seconds on my mid-range GPU.

There's definitely two sides to this, as with most anything.

We are now at the point that you should not blindly trust photographic evidence without additional context or other verifiable measures!