Skip to content

Latest commit

 

History

History
435 lines (323 loc) · 11.3 KB

README_OSC.md

File metadata and controls

435 lines (323 loc) · 11.3 KB

OSC control of Sonic Visualiser

Sonic Visualiser can be controlled remotely using the Open Sound Control protocol. This facility requires the liblo (Lite OSC) library by Steve Harris and Stephen Sinclair to have been available when Sonic Visualiser was built.

Sonic Visualiser opens a single OSC port on startup, provided that OSC support is compiled in, the --no-osc command-line option has not been supplied, and permission to use the network was not refused by the user the first time Sonic Visualiser was run.

The URL of the OSC port is printed to standard output on startup, or can be read from the About box on the Help menu.

General outline

OSC commands accepted by Sonic Visualiser take the form:

<scheme>://<host>:<port>/<method> [<arg> ...]

For example, osc.udp://localhost:12654/play 2.0 will play the current session from time 2.0 seconds.

Methods that manipulate panes or layers act on the currently selected pane or layer. Use the setcurrent method to choose the right target for subsequent such methods.

OSC clients and invocation

If you need an OSC client, there is a small program in the svcore library at svcore/data/osc/sv-osc-send.c that sends an OSC method and arguments to a given URL -- this is not specific to SV but will work with it. To compile that program you should only have to run

$ gcc sv-osc-send.c -o sv-osc-send -llo

provided you have liblo installed.

Then there is a small shell script in the same directory, called sv-command, that provides a basic command shell for Sonic Visualiser. Start SV first, then sv-command should find its OSC port from the system process table when you start it. For example:

$ PATH=.:$PATH ./sv-command      # Set PATH so it can find sv-osc-send
/open snare_hex.wav
/add spectrogram
/set layer Colour Sunset
/play
/quit
$

You can also start Sonic Visualiser directly in "OSC script" mode by supplying the --osc-script <filename> command-line option. The argument to this should be the path of a text file containing a series of OSC commands (like those in the example above). These will be executed in succession as quickly as possible. A line that contains only a number, instead of a command, will cause SV to pause for that number of seconds before going on with the script. Lines beginning with # are ignored.

If you supply "-" as the filename argument to the --osc-script option, SV will instead read OSC commands from standard input. So you can do something like

$ rlwrap sonic-visualiser --osc-script - /path/to/filename.wav

to obtain a rather limited Sonic Visualiser OSC read-eval loop.

Method reference

File management

/open <filename>

Open a new file (of type determined by Sonic Visualiser). If it is an audio file, use it to replace the existing main audio file (if any).

/openadditional <filename> 

Open a new file. If it is an audio file, open it in a new pane in addition to the existing audio file (if any).

/recent <n>
/last

Open the <n>th most recent file from the Recent Files menu, counting from 1 for the most recent file opened. "last" is a synonym for "recent 1".

/save <filename>

Save the current session in <filename> as an SV session file.

This action will try to fail rather than overwrite an existing file, but you probably shouldn't rely on that.

/export <filename>

Export the main audio file to <filename> in floating-point extended WAV format. If a region is selected, export only that region.

This action will try to fail rather than overwrite an existing file, but you probably shouldn't rely on that.

/exportlayer <filename>

Export the current layer to <filename>. See /setcurrent for how to change which layer is current. The format to use is determined from the file extension of <filename>:

  • xml, svl - Sonic Visualiser layer file formats
  • mid, midi - MIDI (note layers only)
  • ttl, n3 - Audio Features Ontology RDF/Turtle (only some layer types)
  • csv - Comma-separated text
  • anything else - Tab-separated text

If a region is currently selected, and the file type is CSV, tab-separated, or MIDI, export only that region. Otherwise export the whole layer.

This action will try to fail rather than overwrite an existing file, but you probably shouldn't rely on that.

/exportimage <filename>

Export a PNG image of the contents of the current pane to <filename>. See /setcurrent for how to change which pane is current. If a region is currently selected, export only that region. Otherwise export the whole pane. Currently the export will fail if the area to be exported is more than 32767 pixels wide.

To export to an image of a fixed width regardless of model duration, consider using /zoom fit to zoom the model contents to fit the window width, then optionally zooming in or out a known number of steps, before exporting.

This action will try to fail rather than overwrite an existing file, but you probably shouldn't rely on that.

/exportsvg <filename>

Export a SVG vector graphic of the contents of the current pane to <filename>. See /setcurrent for how to change which pane is current. If a region is currently selected, export only that region. Otherwise export the whole pane.

This action will try to fail rather than overwrite an existing file, but you probably shouldn't rely on that.

Playback control

/jump <t>
/jump end
/jump selection

Jump the playback position to time <t> (in seconds); or to the end of the file; or to the start of the current selection.

/play
/play <t>
/play selection

Start playback. If a time <t> is given, start from that time in seconds. If the word selection is given instead, play the current selection.

/stop

Stop playback.

/loop on
/loop off

Switch playback loop mode on or off.

Panes, layers, and the window

/add <layertype>
/add <layertype> <channel>

Add a new pane containing a layer of the given type, based on the given channel of the main audio file (numbered from 0). If no <channel> is specified, mix all channels. The specified layer is made the current layer, and the new pane becomes the current pane.

Useful values of <layertype> are:

  waveform
  spectrogram
  melodicrange
  peakfrequency
  spectrum
  timeruler

The following <layertype>s are also accepted:

  timeinstants
  timevalues
  notes
  text
  colour3dplot

But they are less useful, as they create empty layers which there is currently no OSC support for editing.

/setcurrent <pane>
/setcurrent <pane> <layer>

Make the given <pane> (a number counting from 1 for the topmost pane) and optionally the given <layer> on that pane (a number counting from 1 for the "frontmost" layer) the current pane and layer for subsequent pane and layer operations.

/delete pane
/delete layer

Delete the current pane or layer.

/set <control> <value>
/set pane <control> <value>
/set layer <control> <value>

Set a main window control; a property of the current pane; or a property of the current layer.

Accepted main window controls are:

  • gain - whose values are linear multipliers (i.e. 1.0 == unity gain).
  • speed - takes a value of a percentage change in playback speed, so 100 is the default playback speed, 200 sets double the default speed, and 50 sets half the default.
  • overlays - controls the verbosity level of the text overlays on each pane, from 0 (everything off) to 2 (everything on).
  • zoomwheels - controls whether the zoom wheels are displayed (1) or not (0).
  • propertyboxes - controls whether the property boxes are displayed (1) or not (0).

For pane and layer properties, the control name is the displayed name of the given property (though you may use "-" or "_" in place of any spaces in the name if it's easier for you). The value may be the displayed value or underlying integer for the property.

Some examples:

    /set pane Global-Scroll off
    /set pane Follow_Playback Scroll
    /set layer Colour Blue
    /set layer Scale-Units dB
    /set layer Frequency-Scale Log

Note that while you can use "-" or "_" in place of spaces in the property name, you cannot currently do so in the value text. If this is a problem for you, you might be able to set the value as an integer instead (all layer properties can be set this way).

/zoom <level>
/zoom in
/zoom out
/zoom fit
/zoom default

Zoom to a given zoom <level>, given in audio sample frames per pixel; or zoom in or out one step from the current level; or zoom so that the whole model duration fits in the current view width; or return to the default zoom level. This method acts on the current pane (it only affects all panes if set to Global Zoom, which is the default).

/zoomvertical <min> <max>
/zoomvertical in
/zoomvertical out
/zoomvertical default

Change the vertical zoom and origin so as to show the value range from <min> to <max> in the vertical scale; or zoom in or out vertically; or return to the default vertical zoom level. The effect of this method is heavily dependent on the current layer.

/resize <w> <h>

Resize the main window to width <w> and height <h> (if the window system permits).

Region selection

/select <t0> <t1>
/select all
/select none

Select the region from times <t0> to <t1> in seconds; or select the whole file; or clear the selection. If there is a layer selected that can be used as a snap guide for the selection, then the selection will be snapped to it (in the same manner as when making selections interactively).

/addselect <t0> <t1>

Make an additional selection (leaving any existing selection in place) from times <t0> to <t1> in seconds.

Feature extraction

/transform <name>

Transform the current main audio file using the named transform. Transforms are named according to the scheme

    type:source:plugin:output

For example, the percussion onset detector from the Vamp example plugin set can be invoked via

    /transform vamp:vamp-example-plugins:percussiononsets:onsets

If the output is omitted, the first is used. Note that you need to use the plugin and output name, not description: in this case percussiononsets rather than "Simple Percussion Onset Detector".

There is not yet any way to run a transform via OSC on any but the main audio file, nor with any but its default parameters, processing block/step size, or channel selection.

Housekeeping

/undo
/redo

Undo the last editing operation; redo the last undone operation. Note that most of the classic editing operations (copy and paste etc) are not controllable via OSC, but undo may still be useful because Sonic Visualiser considers actions such as adding a pane to be undoable editing operations as well.

/quit

Exit the program abruptly without saving.

Missing things

Handy features still missing from the OSC interface include:

  • the ability to run transforms with non-default parameters or starting from different source models
  • the ability to add layers to a pane (without transform)
  • the ability to add panes (and layers) showing any but the main model
  • the ability to set play parameters on a layer/model and show/hide it
  • the ability to import layers
  • a working pane resize
  • the ability to rename a layer