R1000/Flux-Super-Realism-LoRA-i2i

Prompt

Super Realism, Woman in a red jacket, snowy, hyper-realistic portrait

Model description for super realism engine

Image Processing Parameters

Parameter	Value	Parameter	Value
LR Scheduler	constant	Noise Offset	0.03
Optimizer	AdamW	Multires Noise Discount	0.1
Network Dim	64	Multires Noise Iterations	10
Network Alpha	32	Repeat & Steps	30 & 4380
Epoch	20	Save Every N Epochs	1

Comparison between the base model and related models.

Comparison between the base model FLUX.1-dev and its adapter, a LoRA model tuned for super-realistic realism. [ 28 steps ]

However, it performs better in various aspects compared to its previous models, including face realism, ultra-realism, and others. previous versions [ 28 steps ]

Previous Model Links

Model Name	Description	Link
Canopus-LoRA-Flux-FaceRealism	LoRA model for Face Realism	Canopus-LoRA-Flux-FaceRealism
Canopus-LoRA-Flux-UltraRealism-2.0	LoRA model for Ultra Realism	Canopus-LoRA-Flux-UltraRealism-2.0
Flux.1-Dev-LoRA-HDR-Realism [Experimental Version]	LoRA model for HDR Realism	Flux.1-Dev-LoRA-HDR-Realism
Flux-Realism-FineDetailed	Fine-detailed realism-focused model	Flux-Realism-FineDetailed

Hosted/Demo Links

Demo Name	Description	Link
FLUX-LoRA-DLC	Demo for FLUX LoRA DLC	FLUX-LoRA-DLC
FLUX-REALISM	Demo for FLUX Realism	FLUX-REALISM

Model Training Basic Details

Feature	Description
Labeling	florence2-en (natural language & English)
Total Images Used for Training	55 [Hi-Res]
Best Dimensions	- 1024 x 1024 (Default)
	- 768 x 1024

Flux-Super-Realism-LoRA Model GitHub

Repository Link	Description
Flux-Super-Realism-LoRA	Flux Super Realism LoRA model repository for high-quality realism generation

API Usage / Quick Usage

from gradio_client import Client

client = Client("prithivMLmods/FLUX-REALISM")
result = client.predict(
        prompt="A tiny astronaut hatching from an egg on the moon, 4k, planet theme",
        seed=0,
        width=1024,
        height=1024,
        guidance_scale=6,
        randomize_seed=True,
        api_name="/run"
        #takes minimum of 30 seconds
)
print(result)

Setting Up Flux Space

import torch
from pipelines import DiffusionPipeline

base_model = "black-forest-labs/FLUX.1-dev"
pipe = DiffusionPipeline.from_pretrained(base_model, torch_dtype=torch.bfloat16)

lora_repo = "strangerzonehf/Flux-Super-Realism-LoRA"
trigger_word = "Super Realism"   #triggerword
pipe.load_lora_weights(lora_repo)

device = torch.device("cuda")
pipe.to(device)

Trigger words

Trigger words: You should use Super Realism to trigger the image generation.

• The trigger word is not mandatory; ensure that words like "realistic" and "realism" appear in the image description. The "super realism" trigger word should prompt an exact match to the reference image in the showcase.

Download model

Weights for this model are available in Safetensors format.

Download them in the Files & versions tab.