This LoRA was trained on a character dataset (OneTrainer) for the SD-1-5 Phoskia-Edafos model. It is designed to produce a distinctive low-resolution, brute-force black-and-white photographic style.

The resulting images typically feature low resolution, strong contrast, pronounced film-like noise, and a raw, imperfect aesthetic. The style works particularly well for candid-looking portraits and spontaneous, unpolished photographic moments.

Prompt Structure

slr-quality brutebw photo, [your simple prompt], [portrait style/upper body/full body], [simple background], candid moment, (soft focus:1.2)

e.g., slr-quality brutebw photo, a adult Caucasian man, joy smile, tall athletic build, messy dark hair, sharp facial features, pale skin, large expressive eyes, dark waistcoat, fitted trousers, standing, upper body, simple natural setting as background, candid moment, (soft focus:1.2)

Settings

• Sampler: DPM++ SDE

• Steps: 40

• Prompt length: up to ~75 tokens

• LoRA strength: 0.7–1.2 — experiment to find the desired balance

General Prompting Advice

Keep your prompts relatively simple and concise. Avoid describing every detail of the image. Instead, provide only the key visual cues you want to emphasize and let the model handle the rest.

The LoRA tends to work best when given some creative freedom. Less can be more — overly detailed prompts may interfere with the characteristic style and make the results feel less natural.

Try starting with a simple subject description, specify the framing (portrait, upper body, or full body), and provide only a minimal indication of the environment or mood. The model can often fill in the remaining details surprisingly well. Do not include information about color and light phrases.