Files double in size when 'Compressed'

I have managed to get a satisfactory listening experience with similar output levels (SPL). What I do not understand is how come they are twice as large, on average, as the Raw files I started with. I am using the FLAC encoding throughout my project. I assume that the converted file is a collection of 44,100 sample values for each second and a value (in millivolts I suspect) is retained for each of those 44,100 sample values. They are approaching the size of a ‘WAV’ file.
My question is when dealing with a FLAC encoded file, does the encoder actually recreate a new set of values for each interval?
If not then how is it portrayed in the Compressed or Normalized file? Maybe it keeps some of the original values and code to fudge/bridge the rest, does anyone know how this is accomplished?

Perhaps you inadvertently selected 24-bit FLAC. 24-bit FLAC files are much bigger than 16-bit FLAC.

Hi Steve
I attach a picture of the Macro I use to do some machinations on my Audio Files. All of my files are FLAC on the way in and are around 20-30MB. Outgoing they are from 40-50MB each. I have attached a picture of the Macro I use to manipulate the Audio Files. The songs I am using were ripped back several years ago. They are actually 32 Bit on the incoming and the Macro saves them as 24 Bit on Exporting/Saving. Their size is not all that important to me but I really want to know why it grows. It there any good docs on this depth of subject that you are aware of.
I am trying to figure out something about the process of changing the Audio FLAC file and how/if that is similar or the same as the incoming file when no Transcoding is done (same format in as out). I have done some googling but haven’t found a good answer. Thanks…

I have attached a picture of the Macro I use to manipulate the Audio Files.

I didn’t see any pictures in your post.

Compression specifications can be a little wacky because they’re content dependent. It’s not like WAV where if you have a three minute show, it’s always going to give you a certain size file.

We know that Audacity exports include a dither signal that wasn’t there before. Dither is wide-band, carefully tuned, very low volume white noise. It’s put in the export to avoid bit depth distortion which could occur getting the show in and out of Audacity. If your show is very simple and very low volume, the addition of dither right at the end can make a audio compression much harder and demand a higher compression quality.

Try this. Open a show, turn dither off and see what happens to the output file size. Audacity > Preferences > Quality > Dither.


By default, most audio file formats are imported as 32-bit float regardless of the sample format of the file. In other words, a 16-bit FLAC or 16-bit WAV file will create a 32-bit float “track”.

Your original files will not be 32-bit, because FLAC only supports 16 and 24 bit. 24-bit FLAC files are rare, so it is most likely that your original files are 16-bit, which will account for the 50% file size increase.

Hi kozikowski
Although it appeared to attach it, it did not actually.
I may have got it this time. Anyway I didn’t know that any ‘dither’ was used and obviously it would grow the file. I have had a bad cold for a couple of days but I will get back to this and post here in a couple of days. I do appreciate the information. Cheers…

Hi steve
I did not understand the Bit thing. I guess the 16 Bit and 24Bit are integers? or what? (Normalized values maybe?) I would expect most of the math to be done in double precision and as you say 32 Bit float values. I have just started studying the FLAC Codec standard and I am getting a sense of what it is doing. Not what I had believed at all. I intend to get at this next week, as right now I am with cold. I certainly appreciate the assistance…

Uncompressed “PCM” formats (like WAV and AIFF) represent audio as a sequence of numbers. Each number represents the waveform’s amplitude at a specific time, and is called a “sample”.

These numbers (samples) are binary numbers, represented as a series of 1’s and 0’s (“binary digits” or “bits”). The number of “bits” per sample is known as the “bit depth” or “bit format”.

The samples represent the waveform at equally spaced intervals. The number of samples per second is the “sample rate”.

“CD quality” PCM has 2 channels (stereo), and the samples are at intervals of 1/44100 s (44100 samples per second, in each channel), and each sample is represented as a 16-bit integer.