DEV Community: cherry yan

How I Built an AI Avatar Generator That Outputs Perfect Square Images

cherry yan — Wed, 28 Jan 2026 04:52:36 +0000

Why I Built a Square Avatar Generator (and What I Learned)

When building side projects or small tools, you often don’t realize how many tiny problems add friction—until they stack up.

One issue I kept running into was avatars.

Most platforms (GitHub, social profiles, dashboards, comment systems) expect square images.

But real user photos? Rarely square.

Cropping manually works… until you have to do it over and over again.

So I decided to solve it properly.

The Real Problem with “Simple Cropping”

At first glance, this feels like a solved problem:

Upload an image
Crop to 1:1
Done

In practice, it breaks down fast:

Faces get cut off
Important details fall outside the crop
Automatic center-cropping fails on non-standard poses

What I actually needed was:

Face-aware positioning
Consistent square output
No manual tweaking for every image

That’s when I decided to experiment with an AI-based approach instead of traditional cropping.

From Manual Tools to AI-Based Generation

Rather than just resizing or cropping, I explored using AI to:

Detect the subject properly
Reframe the image into a square
Preserve facial proportions and composition

The result is a workflow where you upload a photo and get a clean, square avatar without thinking about margins, alignment, or cut-off edges.

While working on this, I ended up turning it into a small standalone tool:

SquareFace Generator

It’s focused on one thing only—turning photos into square avatars that actually look right.

What I Learned Building This

A few takeaways from this mini project:

“Small” UX problems are often under-engineered

Cropping sounds trivial until quality actually matters.
AI shines when rules break down

Once images stop being predictable, heuristics fall apart fast.
Constraint-first tools are underrated

Doing one thing well is often better than shipping a toolbox.

Final Thoughts

This wasn’t meant to be a startup idea—just a way to remove friction from my own workflow.

But it reminded me how many developer tools start exactly this way: scratching a very specific itch.

If you’re building side projects, I’d highly recommend paying attention to the small annoyances you repeatedly ignore.

They’re often the best ideas hiding in plain sight.

Happy building 👋

I’m building a small tool to help people understand relationship misunderstandings

cherry yan — Wed, 28 Jan 2026 04:49:30 +0000

Hi DEV 👋

I’m currently building a small side project called DecodeYou.

The idea came from personal experience.
I’ve been in a long-distance relationship, and even when both people care a lot,
misunderstandings still happen — not because of bad intentions,
but because we interpret situations differently.

For example:
the same action can feel “reasonable” to one person,
and deeply hurtful to the other.

Instead of giving advice or labeling people,
DecodeYou focuses on helping users:

reflect before reacting
notice emotional and communication patterns
separate what actually happened from what we assumed it meant

Right now the project is very early.
I’ve open-sourced the repo to document the ideas and structure as I build.

👉 GitHub repo: https://github.com/cherryyan1987/cherryyan1987

I’d really appreciate any feedback, thoughts, or even skepticism.
Happy to learn from this community.

I built a global GLP-1 telehealth directory (and why it’s harder than it looks)

cherry yan — Wed, 28 Jan 2026 04:45:09 +0000

GLP-1 medications (like semaglutide and tirzepatide) have exploded in demand, but finding legitimate, accessible telehealth providers is still surprisingly hard — especially outside the US.

So I built GLP-1 Patches, a global directory of GLP-1 telehealth providers.

What problem I was trying to solve

Most existing resources are:

US-only
SEO farms or affiliate-heavy
Opaque about pricing, prescriptions, or regions served

If you’re not in the US (or you’re comparing options), the information gap is huge.

What the directory includes

Telehealth providers offering GLP-1 medications
Regions served (US / international)
Prescription model (subscription vs one-time)
Transparency around eligibility and process

No prescriptions. No medical advice. Just structured, comparable information.

The hard parts (that I underestimated)

Regulatory differences across countries
Providers changing terms quietly
Avoiding becoming an affiliate spam site
Balancing “useful” vs “compliance-safe” content

Health + software is a very different beast.

Current state

The project is still early, but already useful for:

Users comparing GLP-1 access options
Builders researching the space
Anyone trying to understand how fragmented this market is

👉 Website: https://glp-1patches.org

Feedback welcome

If you’re building in healthtech, telemedicine, or regulated markets, I’d love feedback:

What would make this genuinely useful?
What would you absolutely avoid adding?
Any red flags you see immediately?