Devin Brugman版本v1.0 (ID: 12633)

Devin Brugman is an American model and social media celebrity with more than 1M followers in Instagram.

Update: Retrained once again from scratch for 140 steps (ended up selectin step 135), I pruned my dataset, leaving it at a total of 15 images after discarding and replacing some pics.

Devin's was my 2nd TI here at CivitAI, if I recall correctly, and it was about time to update it. She's one of my favorite models and I've always wanted to make her justice with a good embedding. In fact, Devin was the fist subject I ever tried to make an embedding of, back in January this year... The attempt was such a disaster that I stopped trying for about a month. If you want to have a good laugh, here are the training images the AI generated while learning Dev's physique. The best part is I let it run up until step 4000. I probably was expecting it to improve at some point :_D. (By the way, the attempt was so bad for 2 main reasons: first, because I used Protogen as the training model instead of standard SD, and second because I used a 512x768 format. Rookie mistakes. ^^)

Appreciate my work? My TIs are free, but you can always buy me a coffee. :)

Curious about my work process? I have summarized it here.

Building a good prompt with my TIs

You're obviously free to experiment, but bear in mind that my TIs are trained with a more or less fixed phrasing, that normally starts with:

"photo of EMBEDDING_NAME, a woman"

So I recommend always starting your prompt like that and then building the rest of the prompt from there. For instance, "photo of beautiful (jg_dev1nbr:0.99), a woman with beautiful hair, hair upsweep updo, as a movie star in a (movie premiere), premiere gala, (near a movie theatre), natural skin texture, (elegant denim overalls), (long sleeve shirt), (suspenders:1.2), 24mm, 4k textures, soft cinematic light, adobe lightroom, photolab, hdr, intricate, elegant, highly detailed, sharp focus, ((((cinematic look)))), soothing tones, insane details, intricate details, hyperdetailed, low contrast, soft cinematic light, exposure blend, hdr, faded, (paparazzi in background), (painted lips:1.1), ((looking straight at the camera:1.1)), (hoop earrings)"

描述:

The model has been trained on about 30 good quality images and seems to be working fine with all of the models I've tried (DreamShaper, Deliberate, Clarity, StablyDiffused, base SD 1.5). Enjoy!

Trained on base SD v1.5

Dataset images: 26

Learning rate: fixed 0.004

Batch size: 1

Gradient: Unknown <-- I've lost this piece of data, but I reckon it probably was 1

Steps: 1000

训练词语: devbrugman