This subject embedding is based on img2img generations I made based on Anime Expo's 2013 program guide:
https://nerdreactor.com/2013/04/22/anime-expos-2013-program-guide-is-all-kinds-of-awesome/

Trained and tested on Anything v3.

What is basically does is "mechanises" your image, adding mecha elements to the image while blending with the rest of the prompt in a pretty cool way (mostly white and red mecha suit because of the training sources, but not always, check the examples).
Trigger words are basic triggers from AUTOMATIC1111's subject_filewords.txt.

It seems to work well with a lot of prompts except for some really predominant tags. For example, it works really well with tags like ((hellsing ultimate style)):
(best quality, masterpiece, unity 8k cg wallpaper, cinematic light, futuristic), a photo of a cool expomech, blue hair, high ponytail, (glowing eyes, 1girl, solo, tight clothes), mecha suit, upper body, close to viewer, messy hair, flowing hair, milf, mature, sharp face, beautiful face, assymetrical bangs, perfect anatomy, inside spaceship, serious face, armored core style, epic, ((hellsing ultimate style))

P.S. If your image becomes overly "mechanized" or resembles a robot too much, you can add "gundam, transformers, robot" tags in the negative prompt.