Converting MusicXML to sound and humming vocals

[HOME AUTOMATION]

I'm sure you're familiar with this...

Smoke on the water by Deep Purple.mid

 

This song already features humming...

MMM.KAR

 

Some FAVS...

Bob Marley - Could you be loved.karJAMMIN.KARMANIAC.KARPEG.KARNEVMYLUV.KARDIRTY_LA.KARYOURSONG.KARYOUAINTSEENNOTHINYET.KARYEARCAT.KARWITCHYWO.KARwildwrl1.karVan Morrison - Wild Night.karTHRILLER.KARstarman.karSTAIRWAY.KARSPINWHEL.KARRod Stewart - Young Turks.karROCKET.KARRock the casbah - Clash.karRIDELIKE.KARPEG.KARpeaches.karNeil Sedaka - Bad Blood.karMICHELLE.KARMadness - Our House.karmacpark.karLONGCOOL.KARLOGICAL.KARLoggins & Messina - Your Mama Don't Dance.karLETYOURL.KARLaura Branigan - Self control (4).karLady Marmalade (Labelle).kar

*Karaoke, .KAR files, can be renamed .MID, if need be.  I use "vanBasco's Karaoke Player".

 

**For a Buddy of mine**  RIKKI.KAR

 

Oops ...I might have put too many.

Edited November 16 by HOME AUTOMATION

[Eric Lafortune]

I get the point! A quick test shows that Musescore reads and plays the MIDI files very well. It can export them as MusicXML. My conversion tools read and convert the results fine, within the technical constraints of the sound/speech chips.

 

I'm experimenting with converting MIDI instruments to speech samples (with ConvertWavToLpc, for use as humming samples with ConvertMusicXmlToLpc), with some funny mixed results. The output at 8 kHz remains lo-fi in any case.

 

For rights-free music, I may have to look at classical music and traditional music, which may be more prevalent as MusicXML sheets. Carmina Burana will only be rights-free in 2036! The open-sourced tunes of Team Fortress in MusicXML format are an interesting exception.

[+OLD CS1]

1 hour ago, Eric Lafortune said:

I get the point!

Sorry.  Was not my intention to have the thread flooded with MIDIs.  I got Java working (had to download OpenJDK since Oracle is uncooperative and my Java 1.8 would not run it,) converted the Pitfall! 2 MIDI → SND → WAV and it came out great.  Now I just need to wrap my head around the SND format, as it is not like the sound list format so I cannot use my tools on it.  I also need to go look back at Stefan's MuseScore/4A.

 

 

[Eric Lafortune]

I've released version 3.1 of the Video Tools. It fixes a few bugs and adds more humming samples -- musical instruments this time. So instead of vocal humming, the speech synthesizer imitates musical instruments: banjo, bass, cello, conga, cymbal, drum, guitar, marimba, piano, steel drum, strings, or trumpet. The instrument samples are inevitably heavily processed: resampled at 8 kHz and compressed with the LPC model of a vocal tract, with quantized coefficients. The speech synthesizer plays the results at 8 kHz and 8 bits precision. I've picked the samples that sounded somewhat acceptable. The banjo seemed the most successful...

 

An example: The Logical Song at 180 bpm as a banjo solo:

 

 

You can find the tools, documentation, and source code on Github. Notably, you can find compiled binaries (for Java 14 or higher) in the release section. The jar contains the new samples as LPC text files. You can quite easily create your own samples of musical instruments with ConvertWavToLpc, and optionally ConvertLpcToText to make small modifications.