Name to Avatar
Back to Blog

Why Name to Avatar?

Where it all started: the story behind Name to Avatar

Sergey Atroshchenko

The idea of Name to Avatar was born when I was working on Mapfolks, that allows you to see and share your approximate location with your team and friends.

I wanted to add demo maps to it to show how it works before someone registers. But for a demo map I needed member data, that is basically a bunch or people, everyone of each having a name, avatar and location. Avatars are not mandatory btw, because Mapfolks also renders initials by default. But having proper avatars makes people on map look prettier and more realistic.

Also, the idea was to add a feature to add virtual member to the map by specifying their names. Generating a proper avatar based on the name made a lot of sense.

So the use case was clear, but I needed to decide should I add it as a feature to Mapfolks or make it a separate product. I don't remember my exact reasoning, but there are a couple of reasons it should be separate:

  1. Mapfolks doesn't use AI, and I did not want to add AI to it just for a single feature.
  2. If I make avatar generation an API, it could be reused in other products.

Avatar API

So I started with a plan. I like to discuss big features or project ideas with ChatGPT and then building a plan for Codex.

The initial name for the project was Avatar API which was not very wise because there is already a product with that name (but I learned about it later).

I built a prototype in a couple of hours and it worked! Proof of concept was successful. I deployed it to Vercel and started posting about it on X and sent it to my friends.

Name to Avatar

It got some positive feedback and I decided to buy a domain at once. "avatarapi" was taken so I ended up with nametoavatar.com.

That's how Name to Avatar was born. It took around 4 hours from the first prompt to a working site.

I used Next.js (TypeScript/React) as usual. There were no storage and no auth in the beginning. Text generation was done by Claude, and image generation was done by OpenAI's gpt-image-1 model.

The basic version worked, but it was just a stateless function that received a name and returned an avatar. It was possible to download the avatar, but it was not store anywhere yet and couldn't be reused.

That was still not enough to produce demo data for Mapfolks. The project required persistent storage.

To be continued...