Remove objects from an image. To create my last comic strip, I needed… | by Val Deleplace | Google Cloud – Community | Jul, 2025

To create my last comic strip, I needed variations of an AI-generated image.

The panels consist of the same scene, with more or fewer people in it. To obtain such variants, I could either :

• Ask for a layered image if the model supports it, and open it in GIMP.
• Or ask for several cutout images of objects and integrate them myself. However, the result may lack proper lighting, shadows, and overall consistency.
• Or generate a full-fledged picture, and then remove some of the objects.

This is what I did to create the comic strip : I prompted for a full classroom with Imagen in Google Cloud Vertex AI Media Studio, and then used inpainting to remove some of the students.

My original prompt:

A drawing of 9 young adult students seated in a classroom. We see their face. In the foreground, we see only part of the back of a female teacher. More than 50% of the image height is above the head of the students.

Result:

After a few iterations of removing students, I was ready to assemble the panels and add speech bubbles in Google Drawings.

If you’re interested in fun paradoxes, see the Will Rogers phenomenon, Simpson’s paradox and The Book of Why.

To know more about inpainting features, have a look at the Edit images overview.