OPINION

The Opposite of AI Music Generation

Why the singer stays the singer.

FAMILIAR LABS RESEARCHALL ARTICLES

ABSTRACT

We're the opposite of AI music generation. We oppose replacing the humans. Song Dubbing carries the song you know into another language in the same singer's voice; only the language changes. This is what AI should be used for: letting my mom understand my favorite songs too.

My mother's songs

My mother sings her songs for me, and I can't understand them.

I understand some Mandarin, but sung, the line is gone before I catch it. And I want her to understand my favorite songs.

The words are the part we miss

We all want to understand the lyrics we listen to.

I'm an audiophile, and I play piano, cello and bass guitar. The lyric is the one part of a recording I miss in another language, and understanding it can turn a line you hummed for years into a confession or a goodbye.

The whole team loves karaoke, and nobody goes to karaoke to hear a machine sing. Learn that no human sang a song you liked and you like it less, as you should (why I believe performers stay human).

The singer stays the singer

GENERATING

  • A new song and a new voice, made from a prompt.
  • It has its place where nobody listens closely: an ad, the music in a store.

DUBBING

  • The song you know, the same singer, the same recording.
  • The voice, the phrasing, the breath and the emotion stay; the lyric changes language.

A song is a performance drawn from a lived life; replace the singer and it is a different song. The artist approves every line before anything is released.

What AI should be used for

Dubbing, not generating.

Until now a song reached another language only by being sung again: by the artist, another singer or a new vocal made for it. With Song Dubbing nothing is sung again; the original performance carries over, and the lips in the music video follow the new words. That is the first true song dubbing.

We built the Multimodal Localization Foundation Model (MLFM) to keep a person whole in another language, and a sung line is the hardest version of that job. As a musician, I listen for the multisyllabic and internal rhymes, the meaning, the genre, the rhythm: if the new line doesn't sit in the pocket where the old one sat, it's the wrong line.

People across cultures, connected

++

We like understanding the song.

Song Dubbing is for record labels and their artists, album by album, music videos included (Song Dubbing). To bring an album into new languages, email the team.

DUBBING, NOT GENERATINGHUMAN TO HUMAN

QUESTIONS

Is Song Dubbing AI music generation?

No. It makes no new song and no new voice; it carries an existing recording into another language in the same singer's voice.

Does the singer have to sing it again?

No. With Song Dubbing nothing is sung again: the original performance carries over, which is why Familiar Labs calls it the first true song dubbing.

Who approves the translated lyrics?

The artist or their team approves every line in every language before release.

REFERENCES

  1. [1]Song Dubbing
  2. [2]Human to Human, Not AI to AI

Multimodal Translation

Familiar Labs builds generationally better dubbing with a Multimodal Localization Foundation Model (MLFM), the first foundation model for multimodal translation: one model decides what to localize across text, audio and visual, and shows you the choices. Your voice and lips, carried into new languages. They're a team of 7 with PhDs, backed by Samsung, RedBird, Gold House. Our mission

BACKED BY

  • Wheelhouse
  • Maximum Effort
  • Stanford
  • Yonsei
  • University of Waterloo