Yamaha FS1R Fseq Editor

By Thor Zollinger, writer/musician/engineer at large.  2017

Fseq's are a series of frames of data which script the settings of the synthesizer.  These allow you to play back short sound sequences, like short spoken sentences.  Using Formants, the synthesizer can speek sentences that sound human.  The factory sequences are not very human, though, their sound quality is rough and robotic.  To correct this, I've programmed a new Fseq Editor for the Yamaha FS1R synthesizer which has much higher tonal quality than the existing ones have.  

My ultimate goal is to create a computer program that uses an Fseq in the FS1R to make the synth sing and speak words.  In order to pull that off I need to create an Fseq with all 44 of the standard English phonemes in it.  The existing Fseq Editors just don't do what I need, so I programmed my own.

Here are a few Screen Shots:

http://javelinart.com/images/ThScreen1.jpg   http://javelinart.com/images/ThScreen3.jpg  http://javelinart.com/images/ThScreen4.jpg  http://javelinart.com/images/ThScreen5.jpg  http://javelinart.com/images/ThScreen6.jpg

This is a mature Beta version, so please email me with the bugs you hit in the code so I can fix them (email button in the lower right corner).

Rather than spend time coding wavelet transforms for spectrum analysis I've used the export results from both Audacity and Spectrogram16 instead, and focused my efforts on creating more accurate formant sequences.  If you listen to the clips I've linked below, you should be able to identify who the clip is spoken by.  I've gone for tonal accuracy.  The other fseq editors out there are great, but their timbre is too synthetic and scratchy for my purposes.  I've included Spectrogram16 in the download, since it's getting harder to find it anymore.

  FS1R Stubborness of Dwarves clip   FS1R Ave Maria clip clip   FS1R MMMMMMMM clip 

Yes, these clips are recorded fseq's played back on the FS1R!  Not perfect, but pretty close to the original.  Not bad for using formant's.  The first one is obviously Gandalf, the second is Christina Aguilera. I picked two entirely different types of recordings to sequence, one is a breathy whisper to target the unvoiced operators, the second is a harmonic voice with higher quality timbre.  The third is a singer humming a single note.

I've also tried to convert instrument samples to formants using the editor, but it takes a lot of fiddling with the formant frequencies to get it to work right.  Voices are much easier.  Instruments are better duplicated using the guidelines in my FM Programming Guide on my other webpage.

You can download the Javelin Fseq Editor and manual using the button below.  It runs on Windows XP, Vista, and Windows 7-10.  I don't have a Mac compiler for Delphi, sorry guys.

ver 1.2  I've updated the editor with buttons to save and load pieces of fseq's and I made a couple of other minor upgrades.  All of the features I was intending on installing are there.  Unless you guys find some bugs, the editor is complete.  I also fixed a bug in the file loading routines so it doesn't get stuck on .syx files any more.

ver 1.3  I updated the Editor to extend the maximum sample size from 9 seconds to 99 seconds, it just wasn't long enough for what I needed it to do.  I also made some updates to the Manual, and I've included Spectrogram16 in the download.

Download  Javelin FS1R Fseq Editor v1.3 10.6 MB