Skip to contents

LongSound

Source: R/longsound-wrapper.R
LongSound.Rd

Represents a Praat LongSound object: an audio file that stays on disk instead of loading into memory.

Arguments

.xptr

Not for direct use. External pointer to the underlying C++ LongSound object; set internally when a method returns a new LongSound.

Value

A LongSound object.

Details

A regular Sound loads the whole recording into RAM, which can exceed available memory on a multi-hour field recording, or just crowd out everything else you're working on. LongSound opens the file, reads its header, and streams samples on demand, so you can check duration, inspect amplitude, or pull out short clips from a file far bigger than memory allows. Use extract_part() whenever you need actual waveform data for pitch, formant, or intensity analysis: it reads just the requested time window and hands back a normal Sound.

Usage


ls <- LongSound$open("recording.wav")
part <- ls$extract_part(10, 15)   # 5 seconds as a Sound

Query methods

  • get_duration() - total duration in seconds

  • get_start_time(), get_end_time() - time domain in seconds

  • get_sample_rate() - sampling frequency in Hz

  • get_number_of_samples() - number of samples per channel

  • get_number_of_channels() - number of channels (1 = mono, 2 = stereo)

  • get_file_path() - path to the underlying audio file

  • get_dx() - sampling period in seconds (1 / get_sample_rate())

  • get_x1() - time of the first sample

Time/sample conversion

  • get_time_from_sample(sample) - time, in seconds, of a given sample index

  • get_sample_from_time(time) - index of the sample nearest a given time

Streaming

  • extract_part(from, to, preserve_times = FALSE) - read a time range from disk and return it as a Sound. Set preserve_times = TRUE to keep the extracted Sound's time domain aligned with the original file instead of starting it at time 0.

  • have_window(from, to) - TRUE if that time range is already held in the internal read buffer, so the next extract_part() over it will be fast

  • get_window_extrema(from, to, channel = 1) - minimum and maximum amplitude in a time range, without building a Sound

Save

  • save_part(from, to, path, format = "wav") - write a time range straight to an audio file, without holding the whole clip in memory. format is one of "wav", "aiff", "aifc", "flac", or "wav24" (24-bit WAV).

  • save_channel(channel, path, format = "wav") - write a single channel to an audio file

Utility

  • is_valid() - FALSE if the underlying file handle was closed or garbage collected

  • print() - summary of file path, duration, sample rate, and channel count

Streaming through a very large file repeatedly touches the same disk regions; see longsound_get_buffer_size_pref_seconds to tune how much Praat keeps cached between reads.

See also

Sound

Examples

# LongSound streams from disk, so the example writes a small WAV file first
tmp <- tempfile(fileext = ".wav")
Sound$create_tone(frequency = 150, duration = 1.0)$save(tmp)

ls <- LongSound$open(tmp)
print(ls)
#> <Praat LongSound>
#>   File: /tmp/Rtmpjd61Le/file24e337384c50.wav 
#>   Duration: 1.000 seconds
#>   Sample rate: 44100 Hz
#>   Channels: 1 
#>   Samples: 44100 

cat("Duration:", ls$get_duration(), "s\n")
#> Duration: 1 s
cat("Sample rate:", ls$get_sample_rate(), "Hz\n")
#> Sample rate: 44100 Hz

# Extract first 0.5 seconds as a Sound
sound <- ls$extract_part(0, 0.5)

# Convert a sample index to a time and back
t <- ls$get_time_from_sample(100)
ls$get_sample_from_time(t)
#> [1] 100

unlink(tmp)