Open To Create Newgen Facepacks Using AI

[gedmski]

Hello, I made a previous thread about creating Newgen faces using Stable Diffusion AI. As of now, I have been perfecting a process where I can automate face creation, which takes into consideration the name, age, and nationality of players.

The process is simple:
1. I take a webpage file of the search view consisting of the player's UID, name, age, and nationalities.

2. This webpage file will then be turned into a CSV file and fed into my program.

3. The program creates an image every 5-6 seconds. I tested its capabilities by taking 2040 Bundesliga players (713 players total), and it finished in 1hr 35mins.

Some observable limitations:

• The images aren't transparent. I tried using Deeplab to automate the background removal, but it gives subpar results. remove.bg is a great option, but it takes time (manually), and I have to pay a premium to get automated results.
• There is a 0.1% chance that the image may contain hats.
• Computationally expensive. At the moment, I'm using Google Colab to generate these images.

If you're interested in trying my method, feel free to message me (btw it's for FREE!). I'm also open to any suggestions and improvements.

[Semi Skilled Milk]

Incredible stuff.  I figured it was only a matter of time before someone managed to automate this, so well done!

Does this remove the need for NewGan to get the images into the game?

[gedmski]

50 minutes ago, Semi Skilled Milk said:

Incredible stuff.  I figured it was only a matter of time before someone managed to automate this, so well done!

Does this remove the need for NewGan to get the images into the game?

Technically, yes. Instead of mapping pre-generated faces to UIDs, I assigned the faces upon generation. The only tool that you will need is fmXML.