FLUX.1-dev-LoRA-Cinematic-Octane

Cinematic-Octane

<div class="container"> <img src="https://cdn-uploads.huggingface.co/production/uploads/66dcee3321f901b049f48002/3JRNF2xSS3R7CIAV8bnUy.png" width="1024"/> </div>

Cinematic-Octane is a LoRA model meticulously fine-tuned to embody the warm, dreamy color grading prevalent in contemporary cinema. This model captures the essence of modern filmmaking with its lush, saturated hues and soft contrasts, reminiscent of films that evoke a sense of nostalgia and fantasy. It excels in enhancing the visual depth of scenes, bringing out the vibrant interplay of colors that define today's cinematic aesthetics.

The model's color palette is characterized by rich, golden tones with greenish hue backgrounds and gentle transitions, which are particularly effective in creating a warm, inviting atmosphere. This is achieved through a careful balance of highlights and shadows, where warm colors like orange and yellow dominate while the shadows are more in blue-green tones, providing a sense of comfort and intimacy. The dreamy effect is further accentuated by subtle halation around highlights and a slight grain, adding texture and depth to the visuals. .

<div class="container"> <img src="https://cdn-uploads.huggingface.co/production/uploads/66dcee3321f901b049f48002/hRtSZ5C9YnhACY7jLx3O-.png" width="1024"/> </div>

Trigger words

You should use cinematic_octane to trigger the image generation.

Inference

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16)
pipe.load_lora_weights('aixonlab/FLUX.1-dev-LoRA-Cinematic-Octane', weight_name='cinematic-octane.safetensors')
pipe.fuse_lora(lora_scale=0.8)
pipe.to("cuda")

prompt = "a gorgeous woman and a man, cinematic_octane"

image = pipe(prompt, 
             num_inference_steps=24, 
             guidance_scale=3.0,
             width=768, height=1024,
            ).images[0]
image.save(f"example.png")

Acknowledgements

This model is trained by AIXON Lab. We release this model under permissions. The model follows flux-1-dev-non-commercial-license.