Spanish vs Portuguese lyric transcription: why dialect matters

Spanish and Portuguese sit close on the map and closer in the ear. Close enough that a generic transcription model trips on which language it is even hearing — and that is before you get to the dialects within each, where the slang changes completely from Puerto Rico to Mexico, or from Rio to Lisbon.

Language detection is the first failure point

The transcriber has to decide the language before it writes a word. For Latin music, es and pt are similar enough that detection errors are common, and a track transcribed in the wrong language is unusable. The biggest lever for accuracy here is the speech model and its language identification — not the cleanup that happens later.

Dialect changes the slang, not just the accent

• Spanish: jangueo (PR) vs morra (MX) vs parcero (CO) vs klk (RD) — same language, different vocabulary.
• Portuguese: Brazilian funk slang (mano, quebrada, role) vs European Portuguese (gajo, fixe, bue).
• A single "Spanish" or "Portuguese" model averages these away; per-region tuning keeps them.

Why Musavox treats them separately

Musavox runs dialect-specific modules per region across Spanish and now Portuguese (Brazilian and European), so the lyric sheet reflects the artist’s actual vocabulary. Code-switching with English is handled as a first-class case, which matters across reggaeton, trap and Brazilian funk alike.