feat: new images generated by SD3

[matthewmayer]

I created 100 female and 100 male images using Stable Diffusion 3 via Replicate, to be used in the upcoming faker.image.personPortrait().

Each image is generated at 512x512, and I write a metadata JSON file with the prompt and settings used. The aim is to document the process, but without necessarily making it easy to regenerate all the images repeatedly, since that is is expensive.

Then I use ImageMagick to create the smaller images at 256, 128, 64, 32 and 16 pixels size.

To get a diverse set of pictures the basic prompt is:
profile picture of a {age}-year-old {man, woman} from {country}

I already did a quick check of all images and removed+replaced around 10% of images which looked unprofessional (more "Tinder" than "LinkedIn") or looked too artificial. If there are any other images you think should be removed, please just write a list e.g. male/72 or female/7 and I can replace those with alternatives.

[ST-DDT]

For some strange reason, I feel like I have seen multiple duplicates while reviewing the 128px images, but I couldn't actually find any duplicates.

[ST-DDT]

Suggested change

	You can use the data in the male/metadata and female/metadata to regenerate images. You can run a model locally, or use a service like Replicate using https://github.com/replicate/replicate-javascript. You will need a Replicate API key, and expect it will cost approximately [$0.035 per image](https://replicate.com/pricing) to generate.
	You can use the data in the male/metadata and female/metadata to regenerate images.
	You can run a model locally, or use a service like Replicate using https://github.com/replicate/replicate-javascript.
	You will need a Replicate API key, and expect it will cost approximately [$0.035 per image](https://replicate.com/pricing) to generate.

[ST-DDT]

IMO this is almost good to go.

@xDivisionByZerox What do you think about the images?

[ST-DDT]

We need a license file.

[matthewmayer]

Since AI outputs are generally expected to be PD should we use https://choosealicense.com/licenses/cc0-1.0/ ?

[xDivisionByZerox]

@xDivisionByZerox What do you think about the images?

The images itself look good enough for me. Thank you for starting this @matthewmayer 🙏

For some strange reason, I feel like I have seen multiple duplicates while reviewing the 128px images, but I couldn't actually find any duplicates.

I definitly felt the same way. Guess there is only so much diversity AI can generate.

[xDivisionByZerox]

Since AI outputs are generally expected to be PD should we use choosealicense.com/licenses/cc0-1.0 ?

Team decision:

CC-0 is fine for us. 👍