Model Description
This is a focused style LoRA trained to capture the specific "hard-light grammar" of 1940s Hollywood studio photography and Tri-X film stock. It emphasizes deep blacks, high-contrast "rim" separation, and dramatic shadow cutting (venetian blinds, alleys) while maintaining organic skin texture.
Use it for classic noir portraits, femme-fatale character shots, detective/underworld scenes, and vintage Hollywood/editorial fashion looks where you want strong chiaroscuro and a period-correct atmosphere. Recommended workflow is to run baseline vs LoRA at a moderate weight (~0.5–0.7) and optionally test a higher range (~0.9–1.1) to see how far you can push the signature lighting before controllability or content leakage becomes noticeable.
Description
The SDXL version of this LoRA was trained against the SDXL base model, and it can be used on SDXL base as a portable style layer for 1940s glamour-noir lighting and femme‑fatale mood. In testing, however, the same LoRA tends to look more photographic and more consistent when paired with certain SDXL fine‑tunes, which can better preserve natural tonal transitions, fabric texture, and anatomy while the LoRA pushes the noir “grammar” (hard key, deep falloff, deep blacks).
In practice, that means SDXL base may occasionally interpret the prompt + noir styling in a more graphic/illustrative direction, while a strong fine‑tuned checkpoint can keep the output anchored in an editorial, studio-photography look. Example fine‑tuned checkpoints that were tested with this LoRA are listed in the reference section (and are recommended starting points if the goal is clean, realistic glamour‑noir rather than stylization).
FAQ
Comments (2)
I love this model! My first test images are fabulous. Any chance there will be versions for FLUX.1 or ZIT?
I am currently working on two version of this LoRA for Illustrious. One airbrushed style for pulp magazines, and one inked style for comics. I have moved away from FLUX to Chroma, and I may create a LoRA for that base model. As far as ZIT, I'm kinda holding out to see when the base model might drop before making any LoRAs. BTW, your test images you posted look awesome! Hope to see more...