~nickbp/soundview

Simple audio voiceprint and analyzer for your computer 
Fix backward flags
Fix up Win/OSX instructions
Improve installer output on Windows

refs

main
browse  log 

clone

read-only
https://git.sr.ht/~nickbp/soundview
read/write
git@git.sr.ht:~nickbp/soundview

You can also use your local clone with git send-email.

#SoundView

Simple voiceprint and analyzer for your computer. Shows what's playing/recording with a high definition analyzer and voiceprint.

Licence: GPLv3

#Build and Run

The code is standard C++11 with the following dependencies:

  • CMake for the build system
  • SFML (specifically SFML-Audio and SFML-Graphics) for interacting with system hardware
  • fftw for FFT of the audio stream

All of these tools are built with cross-platform support in mind, and therefore SoundView should more or less inherit that support.

In theory at least.

#Debian and Ubuntu

Here's the steps to build on an apt-based system, with your choice of compiler:

#GCC
sudo apt-get update
sudo apt-get install build-essential cmake git libsfml-dev libfftw3-dev

git clone git@github.com:nickbp/soundview.git

cd soundview; mkdir -p bin; cd bin
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j4

cd apps
./soundview -h
#Clang
sudo apt-get update
sudo apt-get install clang-3.8 cmake git libsfml-dev libfftw3-dev

git clone git@github.com:nickbp/soundview.git

cd soundview; mkdir -p bin; cd bin
CXX=clang++ cmake -DCMAKE_BUILD_TYPE=Release ..
make -j4

cd apps
./soundview -h

#Fedora and Red Hat

Here's the steps to build on a yum-based system, with your choice of compiler:

#GCC
sudo yum update
sudo yum install gcc-c++ cmake git SFML-devel fftw-devel

git clone git@github.com:nickbp/soundview.git

cd soundview; mkdir -p bin; cd bin
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j4

cd apps
./soundview -h
#Clang
sudo yum update
sudo yum install clang cmake git SFML-devel fftw-devel

git clone git@github.com:nickbp/soundview.git

cd soundview; mkdir -p bin; cd bin
CXX=clang++ cmake -DCMAKE_BUILD_TYPE=Release ..
make -j4

cd apps
./soundview -h

#Windows

These steps assume a Win64 build. These are just what worked for me at the time, on a system running Windows 8.1. YMMV.

#Downloads

The following are prerequisite tools/libraries to building SoundView on Windows. Install/unpack these before proceeding below.

  • 7zip since FFTW comes in .tar.gz format
  • Visual Studio (14 aka 2015)
  • git (or use git included in VS?)
  • CMake (3.5.2)
  • NSIS (2.51: technically optional, used to make install package)
  • SFML (2.3.2: "Visual C++ 14 (2015) - 64-bit")
  • FFTW binaries (3.3.4-dll64)
#Build Steps
  1. Grab the soundview repo with git clone.
  2. Generate .lib files for FFTW:
  • Do the lib /def stuff mentioned in the FFTW instructions using 'Developer Command Prompt'.
  • For a 64bit build (which is what these instructions are doing), include an additional /machine:x64 flag to the specified lib command.
  1. Set up CMake using cmake-gui.exe:
  2. Set source dir to root-level soundview/, binary dir to anything.
  3. Hit Configure. In the window that pops up, select Visual Studio 14 2015 Win64, or whatever's appropriate for you.
  4. After waiting for config to finish, two paths need to be configured:
- `fftw_BASE_DIR` = `path/to/fftw-<ver>-dll64/` - Within the unpacked copy of FFTW downloaded earlier. Contains `fftw.h`, `libfftw3-3.lib`, and `libfftw3-3.dll`.
- `sfml_BASE_DIR` = `path/to/SFML-<ver>-windows-vc14-64-bit/SFML-<ver>/` - Within the unpacked copy of SFML downloaded earlier. Contains dirs named `bin`, `include`, and `lib`.
  1. Hit Configure again and observe that a bunch of new paths are resolved from within these directories.
  2. Hit Generate to create <binary dir>/soundview.sln.
  3. Open Visual Studio and point it to the generated <binary dir>/soundview.sln generated above.
  4. Now you can create the build in Visual Studio in one of two ways: Creating an installer (needs NSIS, see above), or just creating the base build.
  • Build an installer package:
    1. Create a RELEASE build of the PACKAGE project. There will be a lot of warnings since by default this project configures VS to be very pedantic.
    2. Browse to <binary dir>, which should contain SoundView-<ver>-win64.exe.
  • Just build the .exe:
    1. Create a RELEASE build of the soundview-app project. There will be a lot of warnings since by default this project configures VS to be very pedantic.
    2. Browse to <binary dir>/Release/apps/, which should contain soundview.exe. Copy the following libraries into <binary dir>/Release/apps/:
    • <binary dir>/soundview/Release/soundview.dll
    • fftw-<ver>-dll64/libfftw3-3.dll
    • SFML-<ver>-windows-vc14-64-bit/SFML-<ver>/bin/: sfml-audio-2.dll, sfml-graphics-2.dll, sfml-system-2.dll, and sfml-window-2.dll
    1. With the .dll's copied into the same directory as soundview.exe, you should now be able to run it.

If you're just seeing slowly extending vertical streaks (or horizontal after pressing Space), it's a sign that SoundView effectively isn't getting any audio. Check the console output to see what devices are being found, and how much audio they're producing. If devices are producing fewer than several thousand "amplitude units" in the console output log, they're basically not producing audio.

The root problem is that Windows doesn't make it easy to tap into whatever audio is currently playing on the system. Your options are to either just display external audio that's brought in via the microphone jack, or to install software that creates a loopback audio device, like this or this, to create a recordable device that mirrors what is currently playing on the system.

#OSX

These steps have been tried on 10.10.

#Downloads

The following are prerequisite tools/libraries to building SoundView on OSX. Install/unpack these before proceeding below.

#Build Steps
  1. Grab the soundview repo with git clone.
  2. Set up CMake using cmake:
  3. Set source dir to root-level soundview/, binary dir to anything.
  4. Hit Configure. In the window that pops up, select Xcode, or whatever's appropriate for you. Select Unix Makefiles to skip opening XCode and instead just run make -j4 via the commandline to build <builddir>/apps/soundview.
  5. CMake should automatically find the versions of FFTW and SFML that you installed via Homebrew earlier, likely under /usr/local/.... If not, update fftw_BASE_DIR and sfml_BASE_DIR to point to them, and re-attempt Configure.
  6. Hit Generate to create the XCode project.
  7. Start XCode, open <builddir>/soundview.xcodeproj
  • Build the project in XCode, then browse to <builddir>/apps/Debug and run the soundview executable.

By default, SoundView seems to only find the microphone on OSX, even when a USB hardware mixer shows up as a device in "Sound". So OSX may just have poor audio support in SFML, or it may have just been the machine I was borrowing for writing these steps.

#Usage

Keyboard shortcuts, features, and options are exposed through the help displayed by ./soundview -h. Here's some additional information on configuring some of those options.

#Device Selection

By default, soundview automatically selects the device to display by selecting the device that produce the most audio. Once soundview is running, this autodetection may be retriggered by pressing D.

./soundview # autodetect device at startup

This autoselection may be overridden by manually specifying a device by id or by name.

./soundview -l # list devices (or see printed list when autodetection is occurring)
./soundview -d 3 # use device 3 in the list
./soundview -d "Sound Blaster 16" # use device with this name

#Display Options

There are multiple features in play for adjusting the appearance of the visualization. Each may be customized via commandline arguments.

  • -f/--fullscreen Start the display in fullscreen mode.
  • --voiceprint-scroll (px) How much to shift the voiceprint for each rendered frame.
  • --loudness-adjust (0-inf) The charted data is self-adjusting relative to the loudness of the audio stream. This determines how quickly to increase sensitivity during quiet periods.
  • --bass-width (0-900) This value may be increased or decreased to adjust the amount of scaling that's given to bass/mids. By default, bass values are given more width in the display than they would otherwise. This makes bass/mids easier to see, otherwise they're very small relative to higher pitches.
  • --lum-exaggeration (0-100) This setting determines how much to brighten quiet values. Quieter values are difficult to see without some exaggeration.
  • --max-lum (0-inf) The maximum luminosity value to use when coloring louder values. Adjusting this changes how colors are displayed.
  • --analyzer-width (%) How much of the display should be taken up by the spectrum analyzer. Setting this to 0 results in only rendering the voiceprint, while 100 results in only rendering the analyzer.

#Performance Options

The display starts at a fairly high definition which can be adjusted up or down via commandline arguments. In particular, the following can be adjusted to increase or decrease the display quality, with proportional changes to system load.

  • --buckets (#) This is the number of columns to be displayed in the spectrum. This is likely the single flag that's most relevant to performance, and it's tied to --audio-sample-rate in that more columns require more data.
  • --sample-rate (Hz) The rate of the stream to read from the audio device. If this is turned too low, the display will tend to refresh at a slower rate since it will be starved for audio data.
  • --collect-rate (Hz) How frequently the audio device should be polled for data. Ideally this should be at or above the display refresh rate, but it shouldn't otherwise have too much impact on performance.
  • --fps-max (Hz) The frames per second to display at. This should be set to the display refresh rate (usually 60, the default), going beyond this just wastes CPU.