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.<h2 id="heading-105">v19:</h2>If you want to try this LoRA, start playing with these settings and experiment from there:<ul><li>Trigger word in v19: cinematic film still</li><li>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</li><li>Recommended model: Realistic Vision 5.1</li><li>Must combine with a darkness LoRA. I used epiNoiseoffset at 0.5 - 0.7 weight <a target="_blank" rel="ugc" href="https://civarchive.com/models/13941/epinoiseoffset">https://civarchive.com/models/13941/epinoiseoffset</a></li><li>Recommended negative embeddings: BadDream, UnrealisticDream</li><li>Start with CFG 4</li></ul><h2 id="heading-106">v05:</h2>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

Files with SHA256: 0cff9967d0c020ddf56ffdc5fdc2f7bba1814a0603c796e105367b06bd7df2c5