Diffusers How to use Danrisi/AnalogCore_Krea2 with Diffusers:
pip install -U diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("Danrisi/AnalogCore_Krea2", dtype=torch.bfloat16, device_map="cuda")
prompt = "A still frame from degraded VHS footage showing a young woman seated inside a train cabin, captured in a candid, intimate moment. She is holding a plain white mug close to her face and sipping from it while looking directly into the camera, creating a quiet but slightly striking connection with the viewer. A faint suggestion of warmth should be visible through a small amount of steam rising from the mug or through the way she holds it, implying a hot drink such as tea or coffee. She has long, wavy brown hair, loosely styled and held back with a simple black headband. Her skin is fair, and her expression is calm, natural, and slightly thoughtful, as if she has just paused mid-sip after noticing the person filming. She is dressed warmly and casually, wearing a vibrant red scarf wrapped around her neck, layered over a brown coat and a light-colored sweater underneath. Several gold rings and small accessories are visible on her fingers as she grips the mug, adding subtle detail and personality. Her posture is cozy and relaxed, seated comfortably in the train seat with one knee pulled up slightly, giving the composition an authentic, lived-in travel feeling rather than a posed portrait. The interior around her includes a maroon or burgundy-colored train seat and a window beside or behind her. In the window, muted outdoor scenery is faintly visible or softly reflected, blurred by motion and distorted by the glass, reinforcing the sense of travel and a cold or overcast day outside. The overall image should feel like a paused VHS recording rather than a clean photograph. Emphasize strong analog video degradation: horizontal tearing across the frame, segmented scanlines, visible time base error distortion, wavering image stability, analog tracking artifacts, low resolution, soft details, color bleeding, mild ghosting, and uneven sharpness. The tones should feel slightly muted and aged, with a warm-but-faded color palette, subtle luminance noise, and the soft, imperfect texture of old home-recorded tape footage. The final result should feel nostalgic, intimate, and unmistakably like a single still image captured from an old VHS cassette."
image = pipe(prompt).images[0] 
