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
[!WARNING] Trigger words: You should use
Super Realismto 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.