dvPortrait

These models generates portraits! They produce photorealistic, medium length portrait generations with relatively easy prompts, and there's two dedicated versions: one trained on stock photos of men and one on women. Both use the same trigger prompt: "dvPortrait" so put that close to the beginning of the prompt.

Each model was trained using The Last Ben Fast Dream Booth on ~20 high resolution, licensed SFW stock photos, at 768px resolution.

All example images generated using Automatic1111 at 768x768 resolution. They all include prompt and sampling data, and here is an example prompt structure for this model, try adjusting the words in italics for different types of people.

Example prompt:
(dvPortrait style:0.7), (beautiful woman:1.2), (85mm:0.9), f1.8, (skin pours:1.0), messy hair, shadows, dynamic pose, professional portrait photography, subtle details, muted, desaturated, low contrast, (hdr:0.7), canon 5D

Negative:
soft, blurry, out of focus, poor quality, hands, deformed, text, type, watermark, logo, underdeveloped, ((pixelated)), (compressed), naked, overexposed

Settings:
Steps: 15, Sampler: Euler, CFG scale: 7, Face restoration: CodeFormer, Size: 768x768, Model: dvPortrait

If you use the models: please leave a review and post some of the images created.