2014-08-08 19:42 GMT+02:00 Fons Adriaensen <fons@linuxaudio.org>:

On Fri, Aug 08, 2014 at 08:46:54AM +0200, Raffaele Morelli wrote:

> Great tool, I finally leaved it untouched as cumulative probability it's
> enough for the scope*.
>
> My 0.02€: I would suggest to add a cli option for output filename, leaving
> 'ebur128-prob' as default if none is specified

Yes, the cumulative plots look less sexy but will give a better idea of
dynamic range than a histogram. The 'loudness range' as displayed by
ebumeter and printed by ebur128 is actually computed from the (modified)
cumulative plots - details in the paper.

cumulative plot is way better to graphically compare two (or a bunch of) audio files dynamic.

statistician here ;-)

here's a few R lines to generate "sexy" plots, the script is supposed to be ran in the same dir where prob files are

require(ggplot2)

require(reshape2)

# read prob files, a shell script is being used to create and rename prob files as outputted from ebur128

files <- list.files(".", "*.prob$")

# load prob files as tables in a list

data.list <- lapply(files, read.table)

# create an empty matrix

s<-matrix(nrow=751, ncol=length(files))

# loads cumulative probabilities into the matrix

for( i in 1:length(files)) {

s[,i]<- data.list[[i]]$V2

}

# matrix to dataframe and proper columns name

s2 <- as.data.frame(s)

colnames(s2) <- files

s2$dbs <- data.list[[1]]$V1 # add X-axis variable (dB)

# melt the data to a long format

df2 <- melt(data = s2, id.vars = "dbs")

# plot, using the aesthetics argument 'colour'

g<-ggplot(data = df2, aes(x = dbs, y = value, colour = variable)) + geom_line()

plot(g)

Next release will have the histograms as well, it's only a few chars extra
code, and the filename as you suggest.

great