Show HN: Trained an AI model to launch my TikTok app
I just launched Glyp, and I wanted to share the story behind it.
A few months ago, I was running content marketing on TikTok for one of my previous apps.
I tried influencer marketing. Paid $200+ per video. Waited days to get the content. Then crossed my fingers hoping it would perform.
Sometimes it worked. Most of the time, it didn’t.
So I switched to AI tools like Arcads. It helped—but only up to a point.
The problem? I couldn’t scale.
Most tools lock you into a pre-made library of actors. You can’t reuse the same face across different scenes. And those same actors? They're already used by other brands. Which kills any chance of building a consistent TikTok presence.
I needed something more flexible. Something I could control end to end.
So I spent the next two months building it.
I trained a custom image-to-speech model that lets you upload a photo of an actor, match it with a voiceover, and generate a 15-second video. Under the hood, it uses WAN 2.1, with custom LoRAs trained for facial expressions, realism, and mouth sync.
Getting the actor to actually express emotion was the hardest part. Neutral faces were easy. But getting them to look surprised, annoyed, or confident—on cue—took weeks of training tweaks and failed outputs.
I also built:
- A scene builder for product references, clothing, and camera angles - A Claude-based script generator trained on UGC ad formulas - Natural-sounding voice filters—like lav mic, car, gym, etc.
The goal wasn’t just to generate videos...
It was to make it possible to run a full TikTok account using a single AI actor— Posting daily, testing angles, holding your product—without ever filming anything.
So I launched Glyp.
If you're a marketer, solo founder, or content creator trying to scale UGC without influencers or production teams, you might find it useful.
If you decide to give Glyp a try, you'll get unlimited image generations and 5 video credits to test it out.
I’d love your feedback—technical, product, or just gut reaction. Happy to answer any questions.
– Chris
No comments yet