ffm_loudnorm: Normalize Loudness in an FFmpeg Pipeline

In tidymedia: Media File Preprocessing and Metadata for the 'tidyverse'

Normalize Loudness in an FFmpeg Pipeline

Description

Add FFmpeg's loudnorm (EBU R128) audio filter. It normalizes the input's perceived loudness toward a target integrated loudness, true-peak ceiling and loudness range. The filter compiles to -af, or joins an existing audio filter chain in the order the filters were added. The glossary in vignette("tidymedia") explains media terms such as LUFS and true peak.

Usage

ffm_loudnorm(
  object,
  target_loudness = -23,
  true_peak = -1,
  loudness_range = 7,
  measured_i = NULL,
  measured_tp = NULL,
  measured_lra = NULL,
  measured_thresh = NULL,
  offset = NULL,
  linear = FALSE,
  print_format = NULL
)

Arguments

object	An FFmpeg pipeline (ffm) object created by ffm_files().
target_loudness	The target integrated loudness, in LUFS (a number in -70..-5). The default, -23, is the EBU R128 target.
true_peak	The maximum true peak, in dBTP (a number in -9..0). The default, -1, is the EBU R128 ceiling.
loudness_range	The target loudness range, in LU (a number in 1..50). The default is 7.
measured_i, measured_tp, measured_lra, measured_thresh	Measured input values from an earlier loudnorm analysis pass: integrated loudness, true peak, loudness range and threshold. Give them together for an accurate two-pass (linear) correction. Give these values and offset as one set, or give none of them. NULL, the default, gives single-pass dynamic normalization. These map to FFmpeg's measured_I, measured_TP, measured_LRA and measured_thresh options.
offset	The target_offset (offset gain) that the analysis pass reports. It is part of the measured set (see measured_i). The default is NULL.
linear	A logical. TRUE requests linear normalization (linear=true), which needs the measured values to hit the target precisely. FALSE (the default) leaves out the option, so the single-pass dynamic behavior does not change.
print_format	The format of the measurement report for an analysis pass: "json", "summary" or "none". NULL (the default) leaves out the option. Use "json" for an analysis pass that a program can parse.

Details

This is single-pass (dynamic) loudnorm. The pipeline stays one reproducible command, with no measurement pass. The defaults follow EBU Recommendation R 128 (2014): target_loudness = -23 LUFS and true_peak = -1 dBTP. Loudness is measured per ITU-R BS.1770-4. The default loudness_range = 7 is FFmpeg's own loudnorm default. EBU R128 does not prescribe a single value.

Two filters are added, not one. loudnorm is followed by asetnsamples, which regroups the filtered audio into frames of 4096 samples and does not pad the last one. Dynamic loudnorm resamples to 192 kHz and gives frames of 192000 samples. Some encoders accept whatever frame they are given, FLAC and Vorbis among them. Even those encoders refuse to open at all on frames of 192000 samples.

Value

object with an added instruction to normalize loudness.

References

EBU Recommendation R 128 (2014), Loudness normalisation and permitted maximum level of audio signals; ITU-R BS.1770-4. https://ffmpeg.org/ffmpeg-filters.html#loudnorm

See Also

normalize_audio(), the task function built on this filter.

Other pipeline functions: ffm_batch(), ffm_codec(), ffm_compile(), ffm_concat(), ffm_copy(), ffm_crop(), ffm_drawbox(), ffm_drop(), ffm_files(), ffm_fps(), ffm_hstack(), ffm_jobs(), ffm_map(), ffm_output_options(), ffm_overlay(), ffm_pixel_format(), ffm_run(), ffm_scale(), ffm_seek(), ffm_trim(), ffm_vstack(), print.tidymedia_ffm()

Examples

video <- system.file("extdata", "sample.mp4", package = "tidymedia")
ffm_files(video, "output.mp4") |>
  ffm_loudnorm() |>
  ffm_compile()

tidymedia documentation built on Oct. 11, 2026, 5:08 p.m.