Customizing dynamic spectrograms

dynaSpec: dynamic spectrogram visualizations

Marcelo Araya-Salas & Matthew R. Wilkins

2026-11-06

 

This article shows how to customize dynamic spectrograms, layer by layer, how to highlight and label parts of the sound with annotations (including frequency contours) and captions, and how to compare sounds. The main vignette shows the basic usage of each layer, and the article Themes for dynamic spectrograms shows the themes. As there, the examples create interactive web pages: press play to watch the spectrogram along with the sound, click on it to play from that point, and use the speed slider to slow the sound down. Use save_dynaspec() with a file name ending in .mp4 to create videos instead.

library(dynaSpec)
library(viridis)
library(seewave)

data("scythebill")
data("antshrike")

Some examples use recordings from the package NatureSounds or downloaded from the web (sounds must be ‘Wave’ objects or local ‘.wav’ or ‘.mp3’ files, so recordings are downloaded first).

Layers can be stored in an object and modified by adding more layers. Arguments of ds_spectrogram() and style layers are combined with the previous ones (so only the settings that change need to be supplied), while other panel and player layers replace previous ones of the same kind, and annotations and captions are added. Settings not supplied take the defaults of each style. dynaspec() sets the frequency range shown (flim, in kHz) and the part of the sound used (tlim, in s: a single value takes the first seconds, two values a time interval).

Spectrograms

ds_spectrogram() sets how the spectrogram is computed (window length wl, overlap ovlp, window function wn) and drawn (color palette pal, amplitude levels of the colors collevels, background colbg). Other arguments are passed to seewave::spectro():

ds_scythebill <- dynaspec(scythebill, flim = c(1.5, 5.5)) +
  ds_spectrogram(wl = 200, collevels = seq(-50, 0, 1))

# a longer recording for scrolling spectrograms
ds_antshrike <- dynaspec(antshrike, flim = c(0.5, 5.5)) +
  ds_spectrogram(wl = 256, collevels = seq(-50, 0, 1))

# palette and amplitude range
ds_scythebill + ds_spectrogram(pal = mako, collevels = seq(-100, 0, 1))

Black and white spectrogram with a wider amplitude range (fainter sounds are shown):

ds_scythebill + ds_spectrogram(pal = reverse.gray.colors.1, collevels = seq(-100, 0, 5))

By default the spectrogram background is the color of the lowest amplitude of the palette (dark purple for ‘viridis’), so quiet parts of the sound and the background look the same. colbg sets another spectrogram background:

ds_scythebill + ds_spectrogram(pal = viridis, colbg = "black")

In paged spectrograms the palette can also be the name of a ‘viridis’ palette (e.g. “magma”) or two colors for a gradient, and amp.trans brightens the spectrogram (a nonlinear transformation that collapses the difference between loud and quiet sounds; 1 means no transformation):

ds_scythebill +
  ds_spectrogram(amp.trans = 3, collevels = seq(-35, 0, 1)) +
  ds_paged(t.display = 1)

Panels

Oscillograms and amplitude envelopes can go above or below the spectrogram (position), with their own color (col) and height relative to the spectrogram (height). Power spectra go on the left or right side (side), with a given width (width, as a proportion of the figure):

cols <- mako(10)

ds_scythebill +
  ds_spectrogram(pal = mako, collevels = seq(-100, 0, 1)) +
  ds_oscillogram(position = "above", col = cols[3], height = 0.3) +
  ds_spectrum(side = "left", col = cols[5], width = 0.12)

An amplitude envelope instead of the oscillogram (oscillograms and envelopes show the same information, so they are better not combined):

ds_scythebill +
  ds_spectrogram(pal = mako, collevels = seq(-100, 0, 1)) +
  ds_envelope(col = cols[5], height = 0.3)

With several rows each row shows its own power spectrum, on the same scale:

data("antshrike")

ds_antshrike <- dynaspec(antshrike, flim = c(0.5, 4)) +
   ds_spectrogram(wl = 256, collevels = seq(-50, 0, 1))

ds_antshrike + ds_spectrum(col = cols[5]) +
  ds_static(rows = 2, height = 350)

In scrolling spectrograms the lower panel can be an overview of the whole sound (ds_overview(), with its relative height), an oscillogram or an envelope:

ds_antshrike +
  ds_envelope(col = cols[5]) +
  ds_scrolling(t.display = 2)

Static spectrograms

Static spectrograms are the default style, so ds_static() is only needed to change their settings or to switch back from another style (the last style added is used). ds_static() sets the color of the play line and active buttons (lcol), the size and resolution of each row (width, height, in pixels, and res), the number of rows (rows), whether the play line is shown (play.line) and the frames per second of videos (fps):

ds_scythebill +
  ds_static(lcol = "#E69F00", width = 1200, height = 400, res = 120)

Without the moving line (clicking on the figure still plays the sound from that point):

ds_scythebill + ds_static(play.line = FALSE)

Scrolling spectrograms

ds_scrolling() sets the time window shown (t.display), whether the time axis moves along with the spectrogram (fix.time = FALSE), the play line (lcol, lty, lwd), the initial speed (speed), a delay before the spectrogram starts moving (buffer, in s), looping (loop), the size and resolution of the frames (width, height, res) and the frames per second of videos (fps):

ds_antshrike +
  ds_spectrogram(pal = viridis, colbg = "black") +
  ds_oscillogram(col = "#31688E99") +
  ds_scrolling(
    t.display = 2,
    fix.time = FALSE,
    speed = 0.5,
    lcol = "white",
    lty = 1,
    width = 1000,
    height = 500,
    res = 120
  )

Long-billed hermit song at 1/5 speed, without axes and looping (web pages loop until paused):

data("Phae.long1", package = "NatureSounds")

dynaspec(Phae.long1, flim = c(2, 9)) +
  ds_spectrogram(
    wl = 300,
    ovlp = 90,
    pal = magma,
    collevels = seq(-50, 0, 5),
    colbg = "black"
  ) +
  ds_scrolling(
    t.display = 1.2,
    width = 1000,
    height = 400,
    res = 120,
    speed = 0.8,
    loop = 3
  )

The first 20 s of a northern nightingale wren recording from xeno-canto using a custom color palette:

nightingale_wren <- tempfile(fileext = ".mp3")
download.file(
  "https://xeno-canto.org/518334/download",
  nightingale_wren,
  mode = "wb"
)

custom_pal <- colorRampPalette(c(
  "#2d2d86",
  "#2d2d86",
  reverse.terrain.colors(10)[5:10]
))

dynaspec(nightingale_wren, flim = c(0, 8), tlim = 20) +
  ds_spectrogram(
    wl = 600,
    ovlp = 95,
    pal = custom_pal,
    collevels = seq(-40, 0, 5),
    colbg = "#2d2d86"
  ) +
  ds_scrolling(
    t.display = 3,
    width = 700,
    height = 350,
    res = 100,
    lcol = "#FFFFFFE6"
  )

Spix’s disc-winged bat inquiry call slowed down 16 times (speed = 0.0625, the slowest speed web browsers can play):

data("thyroptera.est", package = "NatureSounds")

# extract one call
thy_wav <- attributes(thyroptera.est)$wave.objects[[12]]

# convert to 16 bits (as the silence added below)
thy_wav <- tuneR::normalize(thy_wav, unit = "16")

# add silence at both sides
thy_wav <- pastew(
  tuneR::silence(
    duration = 0.05,
    samp.rate = thy_wav@samp.rate,
    xunit = "time"
  ),
  thy_wav,
  output = "Wave"
)
thy_wav <- pastew(
  thy_wav,
  tuneR::silence(
    duration = 0.04,
    samp.rate = thy_wav@samp.rate,
    xunit = "time"
  ),
  output = "Wave"
)

dynaspec(thy_wav, flim = c(12, 45)) +
  ds_spectrogram(
    wl = 400,
    ovlp = 95,
    pal = inferno,
    collevels = seq(-40, 0, 5),
    colbg = "black"
  ) +
  ds_oscillogram(col = inferno(10, alpha = 0.9)[3]) +
  ds_scrolling(
    t.display = 0.08,
    width = 700,
    height = 250,
    res = 100,
    lcol = "#FFFFFFE6",
    lty = 1,
    speed = 0.0625,
    loop = 4
  )

Other options of ds_scrolling() are a custom spectrogram function (spectro.call, a call made with base::call()), spectral derivatives (derivative = TRUE, requires the package ‘imager’) and the window length in ms (hop.size).

Paged spectrograms

ds_paged() sets the duration of each page (t.display), the highlight box (highlight.col, highlight.alpha), the cursor (lcol), a title (title), an amplitude color scale (plot.legend = TRUE), the size of the pages (spec.width, spec.height, in inches) and the frames per second of videos (fps). Other arguments set the preparation of the pages (e.g. resample.rate, amp.thresh or filter):

ds_scythebill +
  ds_spectrogram(amp.trans = 3, collevels = seq(-35, 0, 1)) +
  ds_paged(
    t.display = 1,
    title = "Brown-billed scythebill",
    highlight.col = "#2d2d86",
    highlight.alpha = 0.4,
    lcol = "#FFEA46"
  )

The first 12 s of a humpback whale song, with 3 s pages and the frequency range used by these whales (0-700 Hz):

whale <- tempfile(fileext = ".wav")
download.file(
  "http://www.oceanmammalinst.org/songs/hmpback3.wav",
  whale,
  mode = "wb"
)

dynaspec(whale, flim = c(0.1, 2), tlim = 9) +
  ds_spectrogram(amp.trans = 3, wl = 2000, ovlp = 90, collevels = seq(-50, 0, 5)) +
  ds_paged(t.display = 3)

A yellow-throaded toucan song from xeno-canto with a two-color palette:

toucan <- tempfile(fileext = ".mp3")
download.file("https://xeno-canto.org/154168/download", toucan, mode = "wb")

dynaspec(toucan, tlim = c(3.2, 15.2), flim = c(0, 8)) +
  ds_spectrogram(pal = c("white", "black"), collevels = seq(-100, 0, 5)) +
  ds_paged(t.display = 3, highlight.col = "#d1b0ff", lcol = "#7817ff")

Annotations

ds_annotations() highlights parts of the sound with translucent boxes. It takes a data frame with the start and end of each selection (in s), as the selection tables of warbleR and ohun. If it has ‘bottom.freq’ and ‘top.freq’ columns (in kHz) the boxes only span that frequency range, and colors and labels can be taken from its columns:

# note types
nts <- data.frame(
  start = c(0.34, 1.13, 1.44, 1.75, 2.32),
  end = c(1.03, 1.36, 1.68, 2.28, 2.56),
  bottom.freq = c(2.5, 2.4, 2.3, 2.1, 2.1),
  top.freq = c(3.9, 4, 3.8, 4, 3.7) + 0.5,
  labels = c("A", "B", "B", "C", "D"),
  colors = viridis::mako(n = 7)[1:5]
)

ds_scythebill +
  ds_annotations(nts, col = "colors", labels = "labels")

Annotations are also shown in overview spectrograms:

ds_scythebill +
  ds_scrolling() +
  ds_annotations(nts, col = "colors", labels = "labels") +
  ds_overview()

The box borders are set with border (a color or the name of a column with the color of each box), lwd (width) and lty (line type):

ds_scythebill +
  ds_annotations(
    nts,
    col = "colors",
    alpha = 0.1,
    border = "colors",
    lwd = 2,
    lty = "dotted",
    labels = "labels"
  )

type = "spectrogram" colors the spectrogram itself within each selection, with a gradient from the spectrogram background to the color of the selection (as warbleR::color_spectro()), instead of drawing boxes. A narrower amplitude range (collevels) keeps the background noise uncolored:

# house wren male-female duet recording
hs_wren <- tempfile(fileext = ".wav")
download.file(
  "https://github.com/maRce10/example_sounds/raw/refs/heads/main/house_wren_male_female_duet.wav",
  hs_wren,
  mode = "wb"
)

# and selection table
st <- read.csv(
  "https://github.com/maRce10/example_sounds/raw/refs/heads/main/house_wren_male_female_duet.csv"
)

# colors and labels
st$colors <- viridis(10)[c(3, 8)]
st$labels <- c("male", "female")

duet <- dynaspec(hs_wren, flim = c(1, 14)) +
  ds_spectrogram(
    wl = 512,
    pal = reverse.gray.colors.1,
    collevels = seq(-120, 0, 1)
  ) +
  ds_scrolling(t.display = 1.2, loop = 3, width = 1000, height = 500, res = 120)


duet +
  ds_spectrogram(collevels = seq(-60, 0, 1)) +
  ds_annotations(st, col = "colors", type = "spectrogram", labels = "labels", cex = 4)

type = "line" draws a horizontal line from the start to the end of each selection, at the frequency set with line.freq (in kHz, or the name of a column with the frequency of each line), as is common in figures of duets. Lines are thick and opaque by default (lwd = 8, alpha = 1), and labels are shown above them:

# one line height for each individual
st$line.freq <- c(12, 11.2)

duet +
  ds_annotations(
    st,
    col = "colors",
    type = "line",
    line.freq = "line.freq",
    labels = "labels", 
    cex = 4
  )

Selection tables made with warbleR can be used directly. When they have selections of several sound files, only those of the sound shown are used (matching the file name), or the sound file is chosen with sound.file (e.g. for ‘Wave’ objects):

data("lbh_selec_table", package = "warbleR")
data("Phae.long1", package = "NatureSounds")

dynaspec(Phae.long1, flim = c(1, 10)) +
  ds_spectrogram(wl = 300) +
  ds_annotations(lbh_selec_table, sound.file = "Phae.long1.wav", labels = "selec", col = "tomato")

Frequency contours

Annotations can also show frequency contours, drawn over the spectrogram within each selection. With contour = "dominant" or contour = "fundamental" they are measured on the sound, within the frequency range of the selection (bottom.freq and top.freq, if available) or the frequency limits of dynaspec(). contour.col and contour.lwd set the color and width of the lines:

st <- data.frame(
  start = c(0.34, 1.13, 1.44, 2.31),
  end = c(1.03, 1.36, 1.68, 2.55),
  labels = c("long note", "arched", "arched", "arched"),
  colors = c("#4C78A8", "#B279A2", "#B279A2", "#B279A2")
)

ds_scythebill +
  ds_annotations(
    st,
    col = "colors",
    labels = "labels",
    contour = "dominant",
    contour.col = "#E69F00",
    contour.lwd = 3
  )

Contours can be shown without boxes with alpha = 0, and work in all styles. contour.args sets how they are measured: the window length (wl, by default that of the spectrogram), overlap (ovlp), amplitude threshold below which no frequency is measured (threshold) and smoothing (smooth):

ds_scythebill +
  ds_annotations(
    st,
    alpha = 0,
    contour = "dominant",
    contour.col = "#E69F00",
    contour.args = list(threshold = 10, smooth = 9)
  ) +
  ds_scrolling(t.display = 1, width = 700, height = 300, res = 90)

Dominant frequency contours jump when another frequency becomes the loudest one (e.g. in fast trills), and fundamental frequency contours are coarse for high-pitched sounds at low sampling rates, as in this recording.

Contours measured elsewhere can be supplied as a list with one element per selection: a data frame with the time (in s, as start and end) and frequency (in kHz) of each point, a vector of frequencies evenly spaced along the selection, or NULL for no contour. Supplied contours are drawn as they are, and NA frequencies leave gaps in the line:

my_contours <- list(
  data.frame(
    time = c(0.34, 0.45, 0.5, 0.8, 1.03),
    freq = c(3.7, 3.2, 2.8, 2.9, 2.65)
  ),
  c(2.6, 3.2, 3.7, 3.3, 3, 2.8),
  NULL,
  c(2.2, 2.7, 3.1, 3.2, 3.1, 2.7)
)

ds_scythebill +
  ds_annotations(
    st,
    col = "colors",
    labels = "labels",
    contour = my_contours,
    contour.col = "#E69F00",
    contour.lwd = 3
  )

They can also be a table with one row per selection and the frequencies in numeric columns, such as the output of warbleR::freq_ts():

# freq_ts() reads the sound from a file
data("Phae.long1", package = "NatureSounds")
data("lbh_selec_table", package = "warbleR")

sound_dir <- tempfile()
dir.create(sound_dir)
tuneR::writeWave(Phae.long1, file.path(sound_dir, "Phae.long1.wav"))

hermit_st <- lbh_selec_table[lbh_selec_table$sound.files == "Phae.long1.wav", ]

hermit_contours <- warbleR::freq_ts(
  hermit_st,
  type = "dominant",
  wl = 300,
  length.out = 30,
  img = FALSE,
  pb = FALSE,
  path = sound_dir
)

dynaspec(Phae.long1, flim = c(1, 10)) +
  ds_spectrogram(wl = 300) +
  ds_annotations(
    hermit_st,
    alpha = 0.1,
    contour = hermit_contours,
    contour.col = "tomato"
  )

Captions

ds_captions() adds text at a fixed position of the figure (x and y, as proportions of the figure width and height), shown during a time range of the sound (start and end, in s). fading adds fade-in and fade-out effects (in s), and col, cex and font set the color, size and font. Several captions can be shown one after the other, and several caption layers can be added at different positions. The following code labels the frog species calling in a recording from Costa Rica:

# frog recording from GitHub
frogs <- tempfile(fileext = ".wav")
download.file(
  "https://github.com/maRce10/example_sounds/raw/refs/heads/main/CostaRican_frogs.wav",
  frogs,
  mode = "wb"
)

dynaspec(frogs, flim = c(0, 5.5), tlim = c(35.3, 50.5)) +
  ds_spectrogram(
    wl = 512,
    ovlp = 95,
    pal = cividis,
    collevels = seq(-60, 0, 5),
    colbg = "black"
  ) +
  ds_oscillogram(col = "#31688E", height = 1 / 3) +
  ds_captions(
    c("Dendropsophus ebraccatus", "Eleutherodactylus coqui"),
    start = c(35.7, 42.3),
    end = c(40.8, 50.1),
    x = 0.33,
    y = 0.86,
    col = "#FFEA46CC",
    font = 3,
    fading = 0.6
  ) +
  ds_scrolling(
    t.display = 1.1,
    loop = 3,
    width = 1200,
    height = 550,
    res = 200,
    lcol = "#FFFFFFCC",
    lty = 3
  )

Note that times of annotations and captions refer to the original recording, even when only part of it is used (tlim).

Player

ds_player() sets the speeds offered by the speed slider of web pages (the first one is the initial speed; a single speed hides the slider), whether the pitch is kept when slowing down by default and the additional controls shown (buttons). A player with only the play button, for instance to embed in a presentation:

ds_scythebill + ds_player(speeds = 1, buttons = character())

A player with all controls, starting at half speed and looping (the loop button starts pressed when ds_scrolling(loop) is larger than 1):

ds_antshrike +
  ds_player(
    speeds = c(0.5, 1, 0.25),
    buttons = c("time", "loop", "volume", "download", "fullscreen")
  ) +
  ds_scrolling(t.display = 2, loop = 2)

In scrolling spectrograms, buttons = "zoom" adds a slider (with - and + buttons) to stretch or shrink the spectrogram, from 4 times less to 4 times more time than t.display (at most the whole sound). The spectrogram is drawn at twice the resolution so zooming in stays sharp, which makes web pages larger:

ds_antshrike +
  ds_overview() +
  ds_player(buttons = c("zoom", "volume", "time")) +
  ds_scrolling(t.display = 3, fix.time = FALSE)

The download button saves the sound embedded in the web page (named after the sound file or ‘Wave’ object), and the full screen button enlarges the player to the size of the screen (e.g. for presentations; web browsers may not allow full screen in embedded frames).

Comparing sounds

ds_compare() shows two or more sounds one on top of the other and plays them in sequence, from top to bottom. Layers are added as with dynaspec(), and apply to all sounds. gap sets the silence between sounds (in s). Web pages have one set of controls for all sounds: speed, pitch, volume and zoom apply to all of them, the time readout shows the number of the sound being played and its time, the seek bar spans all sounds and the loop button repeats the whole sequence. Clicking on a spectrogram plays that sound from that point, and then the following ones:

data("female_barn_swallow")
data("male_barn_swallow")

ds_compare(female_barn_swallow, male_barn_swallow, flim = c(0, 10), gap = 0.3) +
  ds_spectrogram(wl = 512, collevels = seq(-50, 0, 1)) +
  ds_oscillogram() +
  ds_player(buttons = c("progress", "time", "loop"))

Any number of sounds can be compared. All of them use the same frequency axis (the lowest Nyquist frequency if their sampling rates differ, as in this example) and the same time scale, so a second takes the same width in all of them. same.scale = FALSE fits each sound to the width of the figure instead:

three_songs <- ds_compare(
  female_barn_swallow,
  male_barn_swallow,
  scythebill
) +
  ds_spectrogram(wl = 256, collevels = seq(-50, 0, 1)) +
  ds_static(height = 300)

three_songs

All styles and panels can be used, as long as all sounds use the same style:

three_songs + ds_scrolling(t.display = 1, width = 700, height = 250, res = 90)

three_songs + ds_paged(t.display = 1)

Sounds can also be dynaspec() objects, for settings or layers of a single sound (e.g. its own time range, annotations or captions). Layers added to the comparison apply to all sounds:

swallow_st <- data.frame(start = 0.2, end = 0.75, labels = "first notes")

female <- dynaspec(female_barn_swallow) +
  ds_annotations(swallow_st, labels = "labels")

male <- dynaspec(male_barn_swallow, tlim = c(0, 2.5))

ds_compare(female, male, flim = c(0, 10)) +
  ds_spectrogram(wl = 512, collevels = seq(-50, 0, 1))

Amplitudes are scaled within each sound, so the colors of the spectrograms and the height of the oscillograms do not compare the amplitude of the sounds. Videos (.mp4) and still images (.png) stack those of each sound: in videos, each sound keeps its first frame until it is played, and its last frame afterwards.

Videos

Videos (save_dynaspec() with a file name ending in .mp4) use the same settings as web pages. Videos of scrolling spectrograms render every frame, which can be sped up by using several cores (parallel) and lowering the frames per second (fps):

save_dynaspec(duet + ds_scrolling(parallel = 4, fps = 30), "house_wren_duet.mp4")

Session information

Click to see

#> R version 4.6.2 (2026-10-29)
#> Platform: x86_64-pc-linux-gnu
#> Running under: Ubuntu 26.04.1 LTS
#> 
#> Matrix products: default
#> BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
#> LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/libopenblasp-r0.3.32.so;  LAPACK version 3.12.0
#> 
#> locale:
#>  [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8       
#>  [4] LC_COLLATE=C.UTF-8     LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8   
#>  [7] LC_PAPER=C.UTF-8       LC_NAME=C              LC_ADDRESS=C          
#> [10] LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C   
#> 
#> time zone: UTC
#> tzcode source: system (glibc)
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#> [1] seewave_2.2.4       viridis_0.6.5       viridisLite_0.4.3  
#> [4] dynaSpec_2.0.0.9000
#> 
#> loaded via a namespace (and not attached):
#>  [1] NatureSounds_1.0.5 sass_0.4.10        tiff_0.1-12        jpeg_0.1-11       
#>  [5] stringi_1.8.9      digest_0.6.39      magrittr_2.0.5     evaluate_1.0.5    
#>  [9] grid_4.6.2         RColorBrewer_1.1-3 fastmap_1.2.0      jsonlite_2.0.0    
#> [13] backports_1.5.1    tuneR_1.4.7        gridExtra_2.3.1    httr_1.4.9        
#> [17] purrr_1.2.2        scales_1.4.0       pbapply_1.7-5      warbleR_1.1.38    
#> [21] textshaping_1.0.5  jquerylib_0.1.4    cli_3.6.6          rlang_1.3.0       
#> [25] fftw_1.0-9         cachem_1.1.0       yaml_2.3.12        otel_0.2.0        
#> [29] parallel_4.6.2     tools_4.6.2        checkmate_2.3.4    ggplot2_4.0.3     
#> [33] curl_8.1.0         vctrs_0.7.3        R6_2.6.1           png_0.1-9         
#> [37] bmp_0.3.1          proxy_0.4-29       dtw_1.23-3         lifecycle_1.0.5   
#> [41] stringr_1.6.0      fs_2.1.0           MASS_7.3-66        ragg_1.5.2        
#> [45] pkgconfig_2.0.3    desc_1.4.3         pkgdown_2.2.1      bslib_0.12.0      
#> [49] gtable_0.3.6       glue_1.8.1         Rcpp_1.1.2         systemfonts_1.3.2 
#> [53] xfun_0.62          knitr_1.52         farver_2.1.2       readbitmap_0.1.5  
#> [57] htmltools_0.5.9    imager_1.0.8       igraph_2.3.4       rmarkdown_2.32    
#> [61] signal_1.8-1       compiler_4.6.2     S7_0.2.2