Text-To-Music

 Re: What Searchable Speech Will Do To You

At the coffee shop, they have pre-made bagels in plastic bins with labels on them. The "Plain" and "Everything" bins were right next to each other, which gave me a song title idea, which then led to the idea of "Plain Somethin'" or "Plain Nuthin'", perhaps with an old-timey blues shuffle feel.

 "Back in the day" (less than a mere century ago) Tin Pan Alley songwriters would use titles as a springboard. Usually, there would be a natural music in them, which would inspire a chorus idea.

This is old hat now--in almost the absolute sense of the phrase; Who would want to write that way? Lovers used to sing to each other in musicals and who does that? Opera is still largely esoteric.

"Back in the day" is almost like "Yesterday" when you sing it, spoken on a downbeat. No one says Yes-TER-day, let alone sing it that way. (If you pronounce "back in the day" on an upbeat it begins to sound like "begin the day").

But the written word and spoken language converted to text is a different animal; there are phonemes, oronyms, and Mondegreens that naturally emerge when words are strung together. Typed text, especially when communicating in short bursts, always remains somewhat ambiguous in its emotional tenor, and can sometimes sound coarse or blunt.

As we know, music can restore language when it has been affected by stroke, Alzheimer's, or head injury. It is evidence that the ancient hard-wiring in our brains naturally connects music with language. Machines are completely changing that, at least superficially.

I think this has interesting implications for music. Perhaps this will make speech more prosodic. Prosody adds another layer of information from inflection and will make all languages tone languages.

[11/24/2024: AI can now sing words, but the syllabication is usually wrong].