Contextual image editing models





Lately, I’ve been integrating Nano Banana and Flux Kontext more into my matte painting workflow. These image editing models work great for a number of tasks. From extending environments seamlessly to making lighting and time of day changes. 

I wanted to show a quick workflow of establishing a quick and solid base for a matte painting.

As my plate, I'll use one of my photos from my library:
A temple nestled in a forest canopy.


I want to extend the temple downward, and have it sit on the edge of a cliff in the forest.

Using Gemini's Nano Banana, I drafted a starting prompt.


For a first pass, I'm happy with the temple extension, but I want the rocks to blend in with the forest more.



This conversational approach makes it possible to split a task into steps if you want to focus on a specific step at a time.



Rocks blended in with foliage more:



Next, prompting for some clouds:






Using Guide Lines

As a next step, I want to extend the temple further, but specifically around the cliff side. I drew some loose guide lines for placement. This way, you still have control over your image and what you want to achieve.



The prompt I put into Nano Banana:



Temple extension further down on the right side:



I provided additional guide lines for the temple to be extended lower to the left:



The final pass with all extensions added:




Lighting changes

Flux Kontext works great for lighting changes. For example here, I asked  to change the light direction from the right to the left with a simple prompt:


Lighting changed to coming from the left:



Changing the time of day to sunset:



Result:




Camera changes

Prompting for camera angle changes is a great way to get a seamless environment extension.



Camera pivoted slightly to the left:



Mood variations

This is very helpful, for example, in achieving quick turnarounds during the visual development stages to establish the look of a scene.














































Comments

Popular posts from this blog

DMP plate rework with GenAI

ComfyUI IC-Light relighting exploration

Creating Styleframes with LeonardoAI and InvokeAI