I'm a full-stack developer — I design and ship web products, automate processes, and set up server infrastructure.
This karaoke service grew out of my ecosystem of multimedia and interactive projects (including the Karaoke Online platform and the voice game Game Voice Reverse). The idea for automating it came from a simple practical frustration: finding a decent karaoke version of a rare track, an indie artist, or a fresh release online is nearly impossible. Existing tools either require hours of manual work in a video editor, or produce low audio quality.
I wanted to combine my experience in audio processing, web technologies, and automation to build a tool that turns any audio recording into a finished karaoke video in a couple of minutes.
The service doesn't rely on someone else's pre-built lyrics database. Every karaoke video is built on the fly, fully automatically, by a pipeline of specialized AI models and automation scripts:
A neural network isolates the clean vocal from the instrumental backing track without losing audio quality.
A speech-recognition model transcribes the vocal to text and syncs every word's timing to the music down to the millisecond.
Backend scripts combine the cleaned instrumental with a background video and generate subtitles with dynamic syllable-level highlighting.
Sing your favorite and rarest songs — the ones that never made it into standard karaoke systems or clubs.
Quickly generate quality instrumental tracks with lyrics for rehearsing and practicing timing.
Learn and memorize foreign-language songs by visually matching the audio to the text.
Questions, integration proposals, or bug reports — reach out through any of these channels: