hoho.bg
It started as a Christmas experiment and quickly became magic. I built an AI-powered Santa who speaks directly to children using words chosen by their parents. When my kids saw it, there were no questions—only belief.
Year
2026
Type
side project
Built with
Visual essay
hoho.bg
01
Image / 01
Last Christmas, I wanted to build something different. Not another project that only makes sense to me, but something my kids would actually love, and something my wife wouldn't just glance at and say "nice" to. Around the same time, I'd been getting into AI video generation, partly inspired by a colleague's work.
That's when it clicked: a Santa who talks directly to my kids, saying whatever words I picked as their dad. I threw together an early version and showed it to them. Their reaction was pure amazement. No explanation needed. They just got it, and they believed it.
That moment stuck with me. It's rare to build something that doesn't need a manual or a five minute pitch. When people immediately get what something is and who it's for, that's when it starts feeling like magic.
Technically it wasn't complicated. I used a voice generation API to turn a parent's words into Santa's voice. But the result felt nothing like a tech demo, it felt magical, and for the kids that was already enough.
At some point curiosity got the better of me and I tried making Santa's mouth move in sync with the audio. If he could talk, he should probably move too. I tried a few JavaScript lip sync libraries. They technically worked, emotionally they didn't. The result was so awkward the kids lost interest almost instantly. The voice that impressed them before got overshadowed by a weirdly twitching beard and an unnatural mouth. What felt magical suddenly felt funny, and not the good kind.
That taught me something I keep coming back to: adding features doesn't automatically add value. Sometimes it does the opposite.
That failure pushed me toward a different approach, using large language models and more advanced AI tooling. That's how I found the Wavespeed AI platform, surprisingly easy to use, well organized, with enough practical examples that experimenting felt easy instead of overwhelming. One model stood out: InfiniteTalk. It takes an image and an audio track and produces the most natural lip sync I'd seen. For the first time, Santa didn't just sound real, he looked real while talking too.
That was the moment it crossed a real threshold, from clever demo to something a kid could genuinely believe in.
Once that was working, the flow finally felt right: a parent writes a short message, the product turns it into Santa's voice, that audio gets combined with his image and brought to life, and out comes a short video of Santa saying exactly what the parent wrote, like he's talking directly to their kid.
When my kids saw it for the first time, it was one of those moments that just doesn't need explaining. The magic was already there.
Beyond making my kids happy, this project gave me something I didn't expect, the chance to share my own version of Santa with them, and quietly keep their belief in him going a little longer.
Next project
Keep exploring