|Subject:||Re: [fluid-dev] Programmatic polyphony count detection in FluidSynth|
|Date:||Tue, 5 Dec 2017 12:56:25 +0100 (CET)|
>Is there any way to programmatically detect how many voices a given MIDI song, using a given soundfont, needs?
This patch (for fluidsynth v 1.1.6) is the tool you need: https://github.com/FluidSynth/fluidsynth/issues/152 (See FluidProfile_0003.zip.)
A configure time using cmake, you must enable the "profiling" option. This will add command line interface and the necessary probes inside the library code.
Then using the fluidsynth application console, you only need to use the "profiling commands" to make your measurements (cpu load, numbers of voices, maximum voices
(i.e polyphony) your hardware is capable).
> Message du 04/12/17 17:54
> De : "Bartosz Nowotny" <address@hidden>
> A : address@hidden
> Copie à :
> Objet : [fluid-dev] Programmatic polyphony count detection in FluidSynth
>BartoszCheers,Hello,I am using FluidSynth on Android. For performance reasons, I am trying to keep the synth.polyphony setting as low as possible while retaining good sounding audio. I have a number songs and soundfonts. I have set up FluidSynth to use 64 simultaneous voices via synth.polyphony and that works for most song and soundfont combinations. Nevertheless, some songs clearly run out of available voices and do not sound right. Is there any way to programmatically detect how many voices a given MIDI song, using a given soundfont, needs? I am trying to strike a balance between performance and good sounding audio. If I pick too little simultaneous voices then I risk some songs not sounding right (or I will have to get rid of them at all), if I pick too many simultaneous voices then I risk sound glitching and playback problems on lower end devices. Knowing how much voices my songs and soundfonts need is the key to getting that balance just right.
fluid-dev mailing list
|[Prev in Thread]||Current Thread||[Next in Thread]|