Kontext-Bottom-Up-View

The Kontext-Bottom-Up-View is an experimental adapter for black-forest-lab's FLUX.1-Kontext-dev, designed to transform scenes into a bottom-up perspective, preserving accurate depth, scale, and lighting direction to enhance overall realism. The model ensures that background elements such as sky or ground surfaces adjust naturally to the new viewing angle, maintaining coherent geometry, texture consistency, and visual balance. It was trained on 800 image pairs (400 start images and 400 end images) to achieve precise, geometry-consistent bottom-up scene generation.

[photo content], recreate the scene from a bottom-up perspective. Preserve accurate depth, scale, and lighting direction to enhance realism. Ensure the background sky or floor elements adjust naturally to the new angle, maintaining authentic shadowing and perspective.

You modified the prompt, altering its properties and subjective elements. Note: this is an experimental adapter and may contain artifacts.

---

Sample Inferences : Demo

---

Parameter Settings

Setting	Value
Module Type	Adapter
Base Model	FLUX.1 Kontext Dev - fp8
Trigger Words	[photo content], recreate the scene from a bottom-up perspective. Preserve accurate depth, scale, and lighting direction to enhance realism. Ensure the background sky or floor elements adjust naturally to the new angle, maintaining authentic shadowing and perspective.
Image Processing Repeats	45
Epochs	24
Save Every N Epochs	1

Labeling: DeepCaption-VLA-7B(natural language & English)

Total Images Used for Training : 800 Image Pairs (400 Start, 400 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	3500 & 400(warm up)
Sampler	euler

---

Trigger words

You should use [photo content] to trigger the image generation.

You should use recreate the scene from a bottom-up perspective. Preserve accurate depth to trigger the image generation.

You should use scale to trigger the image generation.

You should use and lighting direction to enhance realism. Ensure the background sky or floor elements adjust naturally to the new angle to trigger the image generation.

You should use maintaining authentic shadowing and perspective. to trigger the image generation.

Download model

Download them in the Files & versions tab.