# Playing with ComfyUI Using Online Resources

I stumbled upon this very detailed [article](https://www.ilsilfverskiold.com/articles/learn-advanced-image-generation-child-drawings) on how to convert children's drawings into Comfy AI. It explained the various components of the image recognition and generation process, and even had an existing json workflow that we could easily import into our own ComfyUI instance.

It was not difficult to follow up from there. I looked at the various `.safetensors` files that are loaded for the workflow, and merely googled the exact names. Then, it was easy to google where to place those respective files and we are good to go.

In a nutshell, the workflow is about identifying what had been drawn, getting AI to describe the drawing, and then using various image generation components to create an AI depiction.

I got my son into the activity first by showing him the blog. The moment he got home he started drawing and finished before I completed the set up. Here is the original drawing and the final outcome:

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324677617/8fdecd86-3e9c-43b5-8e25-93329e7ce0e4.jpeg align="center")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324710307/552fe7af-d9be-410f-8560-17aad6bfdbf0.png align="left")

Next, I got both my children to draw (including myself).

My son started off with a small drawing. It didn't render too well considering it isn't exactly a frog. But we figured it might need to be of a larger size for it to be able to describe the drawing better.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324716654/cf3a863f-e3eb-4f05-a05f-84ee704c3587.jpeg align="left")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324724155/fc880f0c-5375-44de-b525-9c9bc4590b76.png align="left")

My daughter's simple stick figure rendered quite nicely.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324730260/ad6f0646-89e3-40d8-9af1-7657be5440a6.jpeg align="left")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324735327/ed878e09-703e-411f-b395-bfb9b591d278.png align="left")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324764868/0c89e3cd-59e5-4312-8394-0cb62f952639.jpeg align="left")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324768602/0696bbe7-5b39-4069-a196-1113d56fd449.png align="right")

Knowing that he should draw it larger, my son did another drawing. This time larger and with more colours.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324781474/f8e8805f-6b92-47cb-88ab-37bdb2faeee5.jpeg align="left")

The effect was quite interesting, although it rendered a girl but he was drawing a boy. I couldn't figure out how to amend it yet, because the AI described it as a girl, and I couldn't interfere with the workflow to change the description to a boy. But I could regenerate and hence there are two examples of what was rendered here.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324787077/de2b300b-b18d-4a17-8a69-e2df075294cd.png align="left")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324790866/c0d2f6b1-90e3-4ba8-96f6-70a6b2208d37.png align="left")

I also took the chance to do one of my own. It rendered quite well, but did not get one of the hands and the teeth correct. Nonetheless, it was quite a good quality image.

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324795084/ab9bed2d-d81e-451e-9f26-2da10d73a737.jpeg align="left")

![](https://cdn.hashnode.com/res/hashnode/image/upload/v1755324799978/a5c20361-a4df-4f36-9420-1ecdbf253abe.png align="left")

AI generation isn't perfect for sure, but it has been amazing how far we have gone. In the past, you needed graphic designers to draw your concept professionally, but we can get quite a rough prototype out in less than a minute.