Vamp Test Plugin

h1. About the Vamp Test Plugin

{{>toc}}

The Vamp Test Plugin is a plugin that does no real work, does not analyse its input at all, and always returns the same results. (With one caveat: you should feed it at least 15 seconds of input audio, because it returns features across a 15-second span.)

It has several outputs, which are designed to test the host's handling of various aspects of the Vamp OutputDescriptor SampleType and SampleRate settings. Note that all output features have labels indicating the expected timing and (in some cases) values for the feature.

Think of it as the Vamp plugin equivalent of a TV test pattern or printer test page.

h2. Outputs

The plugin outputs are listed below, with a description and an example of Sonic Annotator's display of the output features. The output comes from Sonic Annotator v0.8 -- use of this plugin turned up a bug in 0.7 which 0.8 fixes!

h3. instants

The *instants* output is a @VariableSampleRate@ output containing ten instants explicitly timestamped at 1.5 second intervals. They are labelled by number and time.

<pre>

"test.wav",0.000000000,"1 of 10 at 0"

,1.500000000,"2 of 10 at 1.5"

,3.000000000,"3 of 10 at 3"

,4.500000000,"4 of 10 at 4.5"

,6.000000000,"5 of 10 at 6"

,7.500000000,"6 of 10 at 7.5"

,9.000000000,"7 of 10 at 9"

,10.500000000,"8 of 10 at 10.5"

,12.000000000,"9 of 10 at 12"

,13.500000000,"10 of 10 at 13.5"

</pre>

h3. curve-oss

The *curve-oss* output is a @OneSamplePerStep@ output containing 20 values from 0 to 0.95 in increments of 0.05. They have implicit timing, so their timestamps will depend on the process block size. The expected times are calculated by the plugin and returned in the feature labels.

<pre>

"test.wav",0.000000000,0,"1 of 20: 0 at 0"

,0.023219954,0.05,"2 of 20: 0.05 at 0.023"

,0.046439909,0.1,"3 of 20: 0.1 at 0.046"

,0.069659863,0.15,"4 of 20: 0.15 at 0.069"

,0.092879818,0.2,"5 of 20: 0.2 at 0.092"

,0.116099773,0.25,"6 of 20: 0.25 at 0.116"

,0.139319727,0.3,"7 of 20: 0.3 at 0.139"

,0.162539682,0.35,"8 of 20: 0.35 at 0.162"

,0.185759637,0.4,"9 of 20: 0.4 at 0.185"

,0.208979591,0.45,"10 of 20: 0.45 at 0.208"

,0.232199546,0.5,"11 of 20: 0.5 at 0.232"

,0.255419501,0.55,"12 of 20: 0.55 at 0.255"

,0.278639455,0.6,"13 of 20: 0.6 at 0.278"

,0.301859410,0.65,"14 of 20: 0.65 at 0.301"

,0.325079365,0.7,"15 of 20: 0.7 at 0.325"

,0.348299319,0.75,"16 of 20: 0.75 at 0.348"

,0.371519274,0.8,"17 of 20: 0.8 at 0.371"

,0.394739229,0.85,"18 of 20: 0.85 at 0.394"

,0.417959183,0.9,"19 of 20: 0.9 at 0.417"

,0.441179138,0.95,"20 of 20: 0.95 at 0.441"

</pre>

h3. curve-fsr

The *curve-fsr* output is a @FixedSampleRate@ output which contains 10 implicitly timed features with values from 0 to 0.9 in increments of 0.1. 

This output has a SampleRate of 2.5, so the features should appear at 0.4 second intervals.

<pre>

"test.wav",0.000000000,0,"1 of 10: 0 at 0"

,0.400000000,0.1,"2 of 10: 0.1 at 0.4"

,0.800000000,0.2,"3 of 10: 0.2 at 0.8"

,1.200000000,0.3,"4 of 10: 0.3 at 1.2"

,1.600000000,0.4,"5 of 10: 0.4 at 1.6"

,2.000000000,0.5,"6 of 10: 0.5 at 2"

,2.400000000,0.6,"7 of 10: 0.6 at 2.4"

,2.800000000,0.7,"8 of 10: 0.7 at 2.8"

,3.200000000,0.8,"9 of 10: 0.8 at 3.2"

,3.600000000,0.9,"10 of 10: 0.9 at 3.6"

</pre>

h3. curve-fsr-timed

The *curve-fsr-timed* output is a @FixedSampleRate@ output in which the features have explicit timestamps. The features have values from 0 to 0.9 in increments of 0.1. They are not, however, equally spaced.

According to the "Vamp Plugin Programmers' Guide":http://vamp-plugins.org/guide.pdf, "The host may round the timestamp according to the sample rate given in the output descriptor's sampleRate field". In my view this is the right thing for a host to do, although it's technically optional. The guide doesn't say whether timestamps should be rounded down or to the closest sample; this plugin assumes rounding to the closest sample. It also isn't clear whether a half-way point should round down or up; I would expect up.

This output tests the basics of this behaviour by providing features at timestamps that do not lie on sample boundaries, along with a label indicating what the host would be expected to snap them to (always the nearest sample boundary).

The output has a SampleRate of 2.5, so the features should be snapped to 0.4 second boundaries.

Thus:

<pre>

"test.wav",0.000000000,0,"1 of 10: 0 at 0 snap to 0"

,0.000000000,0.1,"2 of 10: 0.1 at 0 snap to 0"

,0.000000000,0.2,"3 of 10: 0.2 at 0.166 snap to 0"

,0.400000000,0.3,"4 of 10: 0.3 at 0.333 snap to 0.4"

,2.000000000,0.4,"5 of 10: 0.4 at 2 snap to 2"

,2.000000000,0.5,"6 of 10: 0.5 at 2 snap to 2"

,2.000000000,0.6,"7 of 10: 0.6 at 2.166 snap to 2"

,2.400000000,0.7,"8 of 10: 0.7 at 2.333 snap to 2.4"

,4.000000000,0.8,"9 of 10: 0.8 at 4 snap to 4"

,4.000000000,0.9,"10 of 10: 0.9 at 4 snap to 4"

</pre>

h3. curve-vsr

The *curve-vsr* output is a @VariableSampleRate@ output which contains 10 explicitly timed features at 0.75 second intervals with values from 0 to 0.9 in increments of 0.1. 

<pre>

"test.wav",0.000000000,0,"1 of 10: 0 at 0"

,0.750000000,0.1,"2 of 10: 0.1 at 0.75"

,1.500000000,0.2,"3 of 10: 0.2 at 1.5"

,2.250000000,0.3,"4 of 10: 0.3 at 2.25"

,3.000000000,0.4,"5 of 10: 0.4 at 3"

,3.750000000,0.5,"6 of 10: 0.5 at 3.75"

,4.500000000,0.6,"7 of 10: 0.6 at 4.5"

,5.250000000,0.7,"8 of 10: 0.7 at 5.25"

,6.000000000,0.8,"9 of 10: 0.8 at 6"

,6.750000000,0.9,"10 of 10: 0.9 at 6.75"

</pre>

h3. grid-oss

The *grid-oss* output is a @OneSamplePerStep@ output containing 20 consecutive features of 10 values each, with values ranging from 0 at "bottom left" to (nearly) 1 at "top right".

<pre>

"test.wav",0.000000000,0.0666667,0.1,0.133333,0.166667,0.2,0.233333,0.266667,0.3,0.333333,0.366667,"1 of 20 at 0"

,0.023219954,0.1,0.133333,0.166667,0.2,0.233333,0.266667,0.3,0.333333,0.366667,0.4,"2 of 20 at 0.023"

,0.046439909,0.133333,0.166667,0.2,0.233333,0.266667,0.3,0.333333,0.366667,0.4,0.433333,"3 of 20 at 0.046"

,0.069659863,0.166667,0.2,0.233333,0.266667,0.3,0.333333,0.366667,0.4,0.433333,0.466667,"4 of 20 at 0.069"

,0.092879818,0.2,0.233333,0.266667,0.3,0.333333,0.366667,0.4,0.433333,0.466667,0.5,"5 of 20 at 0.092"

,0.116099773,0.233333,0.266667,0.3,0.333333,0.366667,0.4,0.433333,0.466667,0.5,0.533333,"6 of 20 at 0.116"

,0.139319727,0.266667,0.3,0.333333,0.366667,0.4,0.433333,0.466667,0.5,0.533333,0.566667,"7 of 20 at 0.139"

,0.162539682,0.3,0.333333,0.366667,0.4,0.433333,0.466667,0.5,0.533333,0.566667,0.6,"8 of 20 at 0.162"

,0.185759637,0.333333,0.366667,0.4,0.433333,0.466667,0.5,0.533333,0.566667,0.6,0.633333,"9 of 20 at 0.185"

,0.208979591,0.366667,0.4,0.433333,0.466667,0.5,0.533333,0.566667,0.6,0.633333,0.666667,"10 of 20 at 0.208"

,0.232199546,0.4,0.433333,0.466667,0.5,0.533333,0.566667,0.6,0.633333,0.666667,0.7,"11 of 20 at 0.232"

,0.255419501,0.433333,0.466667,0.5,0.533333,0.566667,0.6,0.633333,0.666667,0.7,0.733333,"12 of 20 at 0.255"

,0.278639455,0.466667,0.5,0.533333,0.566667,0.6,0.633333,0.666667,0.7,0.733333,0.766667,"13 of 20 at 0.278"

,0.301859410,0.5,0.533333,0.566667,0.6,0.633333,0.666667,0.7,0.733333,0.766667,0.8,"14 of 20 at 0.301"

,0.325079365,0.533333,0.566667,0.6,0.633333,0.666667,0.7,0.733333,0.766667,0.8,0.833333,"15 of 20 at 0.325"

,0.348299319,0.566667,0.6,0.633333,0.666667,0.7,0.733333,0.766667,0.8,0.833333,0.866667,"16 of 20 at 0.348"

,0.371519274,0.6,0.633333,0.666667,0.7,0.733333,0.766667,0.8,0.833333,0.866667,0.9,"17 of 20 at 0.371"

,0.394739229,0.633333,0.666667,0.7,0.733333,0.766667,0.8,0.833333,0.866667,0.9,0.933333,"18 of 20 at 0.394"

,0.417959183,0.666667,0.7,0.733333,0.766667,0.8,0.833333,0.866667,0.9,0.933333,0.966667,"19 of 20 at 0.417"

,0.441179138,0.7,0.733333,0.766667,0.8,0.833333,0.866667,0.9,0.933333,0.966667,1,"20 of 20 at 0.441"

</pre>

h3. grid-fsr

The *grid-fsr* output is a @FixedSampleRate@ output which contains 10 implicitly timed features with 10 values each. 

This output has a SampleRate of 2.5, so the features should appear at 0.4 second intervals.

<pre>

"test.wav",0.000000000,0.1,0.15,0.2,0.25,0.3,0.35,0.4,0.45,0.5,0.55,"1 of 10 at 0"

,0.400000000,0.15,0.2,0.25,0.3,0.35,0.4,0.45,0.5,0.55,0.6,"2 of 10 at 0.4"

,0.800000000,0.2,0.25,0.3,0.35,0.4,0.45,0.5,0.55,0.6,0.65,"3 of 10 at 0.8"

,1.200000000,0.25,0.3,0.35,0.4,0.45,0.5,0.55,0.6,0.65,0.7,"4 of 10 at 1.2"

,1.600000000,0.3,0.35,0.4,0.45,0.5,0.55,0.6,0.65,0.7,0.75,"5 of 10 at 1.6"

,2.000000000,0.35,0.4,0.45,0.5,0.55,0.6,0.65,0.7,0.75,0.8,"6 of 10 at 2"

,2.400000000,0.4,0.45,0.5,0.55,0.6,0.65,0.7,0.75,0.8,0.85,"7 of 10 at 2.4"

,2.800000000,0.45,0.5,0.55,0.6,0.65,0.7,0.75,0.8,0.85,0.9,"8 of 10 at 2.8"

,3.200000000,0.5,0.55,0.6,0.65,0.7,0.75,0.8,0.85,0.9,0.95,"9 of 10 at 3.2"

,3.600000000,0.55,0.6,0.65,0.7,0.75,0.8,0.85,0.9,0.95,1,"10 of 10 at 3.6"

</pre>

h3. notes-regions

The *notes-regions* output is a @VariableSampleRate@ output which contains 10 explicitly timed features at 1 second intervals with values from 0 to 0.9 in increments of 0.1. The features have durations of 1.75 and 0.5 seconds respectively (repeated 5 times).

<pre>

"test.wav",0.000000000,1.750000000,0,"1 of 10: 0 at 0 dur. 1.75"

,1.000000000,0.500000000,0.1,"2 of 10: 0.1 at 1 dur. 0.5"

,2.000000000,1.750000000,0.2,"3 of 10: 0.2 at 2 dur. 1.75"

,3.000000000,0.500000000,0.3,"4 of 10: 0.3 at 3 dur. 0.5"

,4.000000000,1.750000000,0.4,"5 of 10: 0.4 at 4 dur. 1.75"

,5.000000000,0.500000000,0.5,"6 of 10: 0.5 at 5 dur. 0.5"

,6.000000000,1.750000000,0.6,"7 of 10: 0.6 at 6 dur. 1.75"

,7.000000000,0.500000000,0.7,"8 of 10: 0.7 at 7 dur. 0.5"

,8.000000000,1.750000000,0.8,"9 of 10: 0.8 at 8 dur. 1.75"

,9.000000000,0.500000000,0.9,"10 of 10: 0.9 at 9 dur. 0.5"

</pre>

h3. input-summary

The *input-summary* output is a @OneSamplePerStep@ output which produces data derived from the input to the plugin, so as to test whether the host is providing the expected input. Each feature contains one value per input channel. That value consists of the first sample value on that channel plus the total number of non-zero samples on that channel within the associated block of input. ("Non-zero" is determined by comparison against a magnitude threshold which is actually 1e-6 rather than exactly zero.)

Since this output depends on the input, a sample is not given here.

h3. input-timestamp

The *input-timestamp* output is a @OneSamplePerStep@ output which produces single-valued features, whose value is the time in sample frames converted from the timestamp of the corresponding process input block.

Since this output depends on the input, a sample is not given here.