Vidome

VIDOME turns a song and its lyrics into a finished lyric video. Upload the track (MP3, WAV, M4A, FLAC or OGG), paste the words or let it transcribe them, and a speech model plus an alignment pass puts every word on the beat it is sung on — each word with its own in and out point, not line-by-line captions. The alignment keeps only timing from the model, so a misheard transcript can never change the words on screen.

Mistimed words are fixed by tapping them during playback or nudging a line in 50 ms steps; highlighting leads the vocal by 55 ms so it reads as on time. Backgrounds react to the track's waveform in 50 ms buckets. There are 33 themes (11 free, 7 with camera, film grain and light), 14 word-arrival styles, 15 lighting styles and 14 typefaces. One project renders landscape for YouTube and portrait for Reels, Shorts and TikTok; the type rescales and re-wraps for the vertical frame.

Rendering happens in the browser — Canvas in a Web Worker — and is paid in credits: 3 to analyse a track, 8 to 39 to render from 720p to 4K/60, 20 free on signup with no card. Earl wrote the alignment approach, the renderer and the editor. The about page leads with where it struggles, which is a design decision too.