WATCHES

This is a Style/Concept/Object LoRA for generating watches. It includes a number of different watch styles and brands, as well as various case material and band styles. Both a SDXL and SD V1-5 version LoRA are provided. This was an experiment to explore model flexibility within a single class using tagging of sub-styles elements, in this case different watch types and brands.

Along with the trigger word W4tch35, you can use the following to elicit other styles and details:

chronograph, luxury_watch, dress_watch, field_watch, dive_watch, aviator_watch, racing_watch, fashion_watch, sport_watch, pocket_watch.

This model should be inherently SFW, although the dataset included a few risque watch faces, so include NSFW as a negative prompt if you have any concerns.

This model is a SDXL LoRA trained on 176 images using Kohya_ss (https://github.com/kohya-ss/sd-scripts) webui (https://github.com/bmaltais/kohya_ss). Sample images were rendered using A1111 (https://github.com/AUTOMATIC1111/stable-diffusion-webui).

Sample Prompt:

<lora:W4tch35_1-5:1.0>,an advertising photograph of a beautiful (W4tch35:1.2) as a aviator_watch,on a watch stand,with detailed movement,medium shot,tilt shift,bokhe,8k wallpaper,masterpiece,photorealistic,HDR UHD,high contrast,hyper realistic,volumetric lighting,

If you are pleased with the models provided, I would be grateful for your support with a 👍 Like, 🚶 Follow, or ❤ Review. I greatly appreciate your comments, feedback, and images which helps me continue to improve my model workflow and testing.

I am also extremely grateful for tips ☕ if you appreciate my work, (https://buymeacoffee.com/Signalytix), as it helps significantly to offset the cost and effort that I devote to research, create, contribute, and share my work with the community. Suggestions posted to my tips page garner my utmost attention for new models. I currently focus on development of LoRAs and embeddings for SDXL and V1.5.

------------------------------------------------------

Model Card:

Name: W4tch35_1-5.safetensors

Version: 1-SD1-5

Description: Dress watches of varying styles

Trigger Words: W4tch35

Data Set: 176 images (512)

Epochs: 6

Steps: 1056

Batches: 2

Repeats: 4

Network Dim/Alpha: 96/48

Optimizer: Prodigy

Base Model: stable-diffusion-v1-5

------------------------------------------------------

ARTIFICIAL HUMANS ASSOCIATION (AHA) CERTIFICATION:

No WAIFUS were harmed, exploited, or compromised in the training of this model or creation of images.