Type | |
Stats | 628 768 |
Reviews | (71) |
Published | Mar 12, 2024 |
Base Model | |
Trigger Words | cinematic film still |
Hash | AutoV2 0CFF9967D0 |
This LoRA is intended to reproduce the neon-like cinematic look popularized by A24, most notably in the tv show Euphoria.
I made 22 versions of this style LoRA. I'm not happy with the result, but I'm stopping here and sharing the best 2 versions I managed to make (v19 and v05). I was trying to make the LoRA learn interesting compositions, lighting, and colors typical of this cinematic look. Unfortunately the LoRA seems to learn a bit of everything - overlearns things we don't want it to learn, and underlearns things we want it to learn. It is what it is. All the samples are cherry picked.
v19:
If you want to try this LoRA, start playing with these settings and experiment from there:
Trigger word in v19: cinematic film still
Other positive prompt words to try with this to improve the look: reflections, fog, dark, dimly lit, bokeh, dreamy, soft focus, purple, blue, pink, green, woman, man, couple, at distance
Recommended model: Realistic Vision 5.1
Must combine with a darkness LoRA. I used epiNoiseoffset at 0.5 - 0.7 weight
https://civitai.com/models/13941/epinoiseoffsetRecommended negative embeddings: BadDream, UnrealisticDream
Start with CFG 4
v05:
The newer version should be preferred, but for some prompts this old version seems to achieve more pleasing compositions and better contrast in dark images, so I decided to share this old version too.
You can use similarly as v19, but trigger words are different: euphoria style film still, purple hue, blue hue, pink hue, dark, low key, dimly lit, film grain