For the past few months, I’ve been working on a small side project that I finally feel comfortable sharing.
It’s called AnimFlow - https://animflow.io/
At a high level, AnimFlow is a simple idea:
Take a single image - a mascot, a character, a sticker - and turn it into a short AI-generated animation.
Not a cinematic video.
Not a full production pipeline.
Just a clean, short, expressive motion that feels usable.
What it actually does
The current version is very focused.
You upload an image
You describe a small action (e.g. “wave hello”, “idle breathing”)
You get a short animation back
The output is intentionally:
short
loopable
consistent with the original character
It’s designed more for animated characters, stickers, mascots, and lightweight content, not full videos.
Additionally, I think I am giving the best price on the market. Other popular services that I tried charge much more.
Here is the animation of my mascot character, which I generated in my service:
How I’m running it
From a technical perspective, it’s a fairly standard setup:
Django backend + REST API
Async job system for animation generation based on Celery
GPU workers running the actual model, hosting on Runpod.io. Currently, I run optimized Wan2.2, but planning to try / integrate other models.
Next.js Web app as the main client, I have also almost finished the iOS app
The important part is that everything is job-based:
You submit an animation request, it runs asynchronously, and you get the result when it’s ready. In the iOS app, you even get push notifications.
I’m keeping the system simple and modular so I can iterate quickly.
What’s next
Right now, this is very much an MVP.
My focus going forward:
improve motion quality and stability
try new models and experiments
make outputs more “sticker-ready” + try different export formats (GIFs, spritesheets, etc)
I’m also trying to keep this as a small, sustainable project, not overbuild it.
Final thoughts
This is my first time launching something that feels like an actual SaaS product —
even if it’s still early.
I’m not expecting it to explode or anything.
But it’s a good step:
building something end-to-end
putting it in front of users
and seeing what actually sticks
If you’re curious, I’ll be sharing updates as I iterate on it.
