Type | |
Stats | 252 23 87 |
Reviews | (47) |
Published | Feb 2, 2025 |
Base Model | |
Training | Steps: 4,000 Epochs: 17 |
Usage Tips | Clip Skip: 1 Strength: 1.8 |
Trigger Words | G3RTHS3M4N3 |
Hash | AutoV2 2CC953A4B5 |
Please read if you want to get the best results.
This Lora Pack has a few different models designed to be used in conjunction with one another to allow control of the level of image degradation/enhancement.
1) Vintage Horror_1930s Black and White Film: Use the tag "G3RTHS3M4N3" in your prompt. Strength 1.4. When this model is alone, this will produce images that have a level of degradation that matches the 360p training data which comes from several 35mm and 16mm 1930s horror films. If you would like to create images that are less-degraded/more-enhanced use it in conjunction with the Enhancer version, noted below.
2) Vintage Horror_1930s Black and White Film_Enhancer: Use the tag "G3RTHS3M4N3" in your prompt. Use this model at a strength of 0.3-0.9 in conjunction with the non-enhanced version to achieve a more refined, modern, look. This model was trained on a smaller dataset enhanced with Magnifique. If used alone, this will not produce vintage analog effects.
3) Nightmare Creatures: Use the tag "M0N5TRP0RTR41T" in your prompt. This model can be used on its own to create monsters and demons but does not produce vintage analog effects, however, I am including it in this Lora pack because it pairs nicely with the vintage film models when used at a strength of .6
4) Cult Symbols_Caligraphy/Typography: Use the tag "CUL7SYMBL5" in your prompt. This model can be used on its own to create cult-like calligraphy at a strength of 2.0, I am including it in this Lora pack because it pairs nicely with the vintage film models when used at a strength of 1.0.
Recommended settings for Vintage Horror - 1930s Black and White Film:
resolution: 1024x768 - all of the training data for these loras was resized to reflect the 4:3 ratio.
Strength: 1.4 - as low as 0.6 for subtle effects, as high as 1.8 for more intense ones.
Flux Guidance: 2.2 - as low as 2.1 for more realism, as high as 2.5 for more anime.
Max Shift + Base Shift: these affect the size of the grain/texture pattern.
Max Shift: 0
Base Shfit: 2.0 - as low as 0 for smaller grain patterns, as high as 6 for larger grain patterns.
Scheduler + Sampler Combos - these affect the texture rendition:
unipc_bh2 + normal = standard look
euler + simple = less grain
dpm_fast + sgm_uniform = darker
dpm_2 + karras = darker, embraces abstraction
dpm_fast + beta = darker, washed out, motion blur, embraces abstraction
Hearing your feed back and seeing what you create helps me improve future models.
Cheers, and happy Cult-Monster-Demon-Horror Creating.