The Kontext-Unblur-Upscale is an experimental adapter for black-forest-lab's FLUX.1-Kontext-dev, designed to upscale low-quality images to 4K resolution, enhancing sharpness, clarity, and fine details while preserving the original texture, colors, lighting, and natural appearance. The model effectively removes noise, blur, and compression artifacts, ensuring high-fidelity restoration and visually realistic results. It was trained on 1,000 image pairs (500 start images and 500 end images) to deliver precise, artifact-free upscaling performance.
[!note]
[photo content], upscale the low-quality image to 4K resolution, enhancing sharpness, clarity, and fine details while preserving the original texture, colors, lighting, and natural appearance. Remove noise, blur, and compression artifacts without over-smoothing or distorting facial or object features. Ensure realistic depth, balanced contrast, and accurate tones, achieving a high-definition, lifelike result that maintains the integrity of the original image.
You modified the prompt, altering its properties and subjective elements. Note: this is an experimental adapter and may contain artifacts.
Sample Inferences : Demo
Kontext-Unblur-Upscale
Kontext-Unblur-Upscale
Kontext-Unblur-Upscale
Kontext-Unblur-Upscale
Parameter Settings
Setting
Value
Module Type
Adapter
Base Model
FLUX.1 Kontext Dev - fp8
Trigger Words
[photo content], upscale the low-quality image to 4K resolution, enhancing sharpness, clarity, and fine details while preserving the original texture, colors, lighting, and natural appearance. Remove noise, blur, and compression artifacts without over-smoothing or distorting facial or object features. Ensure realistic depth, balanced contrast, and accurate tones, achieving a high-definition, lifelike result that maintains the integrity of the original image.
Image Processing Repeats
50
Epochs
30
Save Every N Epochs
1
Labeling: DeepCaption-VLA-7B(natural language & English)
Total Images Used for Training : 1000 Image Pairs (500 Start, 500 End)
Training Parameters
Setting
Value
Seed
-
Clip Skip
-
Text Encoder LR
0.00001
UNet LR
0.00005
LR Scheduler
constant
Optimizer
AdamW8bit
Network Dimension
64
Network Alpha
32
Gradient Accumulation Steps
-
Label Parameters
Setting
Value
Shuffle Caption
-
Keep N Tokens
-
Advanced Parameters
Setting
Value
Noise Offset
0.03
Multires Noise Discount
0.1
Multires Noise Iterations
10
Conv Dimension
-
Conv Alpha
-
Batch Size
-
Steps
4500 & 400(warm up)
Sampler
euler
Trigger words
You should use [photo content] to trigger the image generation.
You should use upscale the low-quality image to 4K resolution to trigger the image generation.
You should use enhancing sharpness to trigger the image generation.
You should use clarity to trigger the image generation.
You should use and fine details while preserving the original texture to trigger the image generation.
You should use colors to trigger the image generation.
You should use lighting to trigger the image generation.
You should use and natural appearance. Remove noise to trigger the image generation.
You should use blur to trigger the image generation.
You should use and compression artifacts without over-smoothing or distorting facial or object features. Ensure realistic depth to trigger the image generation.
You should use balanced contrast to trigger the image generation.
You should use and accurate tones to trigger the image generation.
You should use achieving a high-definition to trigger the image generation.
You should use lifelike result that maintains the integrity of the original image.
