Size Of Edited Files

My version of Windows is Vista Premium SP2. My version of Audacity is 2.0.3. I downloaded the executable file.

I have a huge music library. I use Audacity a lot. When and wherever I can, I purchase only flac files. They are 2-3 or more times the size they would be if they were mp3. I’m afraid I’ll soon run out of space on my 1 terabyte external HD. Here is an example of my problem:

I had a flac file about 13.1 MB in size. The length was 2:13. Except for one spike in volume (amplitude?) which lasted about 1/10 of a second, the file was of excellent quality. I put the file in Audacity, highlighted the spike and used Amplify effect to incrementally reduce the volume of the spike til I felt it matched the rest of the file. I saved the file as flac.

The results were perfect, but I ended up with a file almost twice the size of the original. On a few occasions, I have had to put an edited file into Audacity a second time. When I do, a file which originally was about 20 MB, is now 70 MB or more. I wouldn’t expect a file to increase so much in size unless I increased Bass and/or Treble or compressed the file. Is there any way to keep exported files from doubling and tripling in size and still maintain all the quality?

I had a flac file about 13.1 MB in size. The length was 2:13.

That’s about right. CDs are about 10MB per minute and FLAC makes a file about 60% of the original size.

If you are not changing the sample rate (kHz) or bit depth, your file size should not change significantly. At higher sample rate, larger bit-depths, or with more channels, a FLAC can be bigger than an uncompressed “CD-quality” WAV file.

You are exporting to FLAC, right?

Dither can make files harder to compress losslessly, so you can end-up with larger files. I don’t think Audacity will add dither in this situation, but just to be sure go to Edit → Preferences → Quality and make sure both dither boxes say “None”.

Download [u]Mediainfo[/u]***** and check the format before & after. You should see something like, “860kbps, 44.1kHz, 16-bits, 2 Channels, FLAC”.

The bitrate (kbps) is the number of bits in one second of sound. If you know the playing time and you know there are 8 bits in a byte and 60 seconds in one minute, you can calculate file size from the bitrate.

For uncompressed files, you can calculate the bitrate if you know the sample rate, but depth and number of channels. And if you know the playing time you can calculate file size. (CDs are 44,100Hz, 16-bits, 2-Channels… 44.1kHz x 16 x 2 x 8 = 1411kbps).

\

  • When you install Medianfo, it tries to install some crapware or some kind of browser “enhancement”, etc. Check the “No”, or “Do Not Accept” box when it asks to install something other than Mediainfo.

Normally FLAC files are about 50 to 60% of the size of an uncompressed WAV file.
MP3s are typically between 10 to 20% of the size on an uncompressed WAV file.

So for a 16 bit 44100 Hz stereo uncompressed WAV file with a file size of say 30 MB (roughly 3 minutes duration), I would expect an MP3 file to be between about 3.5 to 7 MB, and a FLAC version to be around 15 to 18 MB.

That doesn’t make much sense unless you have dramatically increased the sample rate. Even converting from mono to stereo is only likely to double the size.

What audio player do you use to play the files? Does it give you any information about the file format?

I mostly use VLC. I get my file size and format information from Windows. I don’t hide my extensions. Also, when I hover the mouse over the file, a little window opens, showing that same information.

If I have a flac file that sounds sort of flat, I use Audacity to enhance the file to suit my listening preferences due to some hearing loss.To help us determine why I’m getting large files, I put the file below in Audacity and made changes I might normally make.

\

General
Complete name : E:MusicBERT KAEMPFERTBaby Elephant Walk.flac
Format : FLAC
Format/Info : Free Lossless Audio Codec
File size : 15.2 MiB
Duration : 2mn 51s
Overall bit rate mode : Variable
Overall bit rate : 744 Kbps
Album : Bert Kaempfert And His Orchestra
Album/Performer : Bert Kaempfert / Bert Kaempfert
Track name : Baby Elephant Walk
Track name/Position : 14
Performer : Bert Kaempfert
Genre : Instrumental
Recorded date : 1978

Audio
Format : FLAC
Format/Info : Free Lossless Audio Codec
Duration : 2mn 51s
Bit rate mode : Variable
Bit rate : 744 Kbps
Channel(s) : 2 channels
Sampling rate : 44.1 KHz
Bit depth : 16 bits
Stream size : 15.2 MiB (100%)
Writing library : libFLAC 1.2.1 (UTC 2007-09-17)

=======

I increased the Bass by three decibles. I increased the treble by two decibles. It sounded better, but still flat. Using Compress Dynamics 1.2.6, I set the Compress ratio to 0.3 and Compression hardness to 0.3. I leave the Floor preset at -32.0. I leave the Noise gate falloff preset at 0.0. I leave the Maximum amplitude preset at 0.99. After compression, the file sounds much better.

The last thing I did, which often helps me hear instruments I didn’t hear before, is to offset the tracks. I split the stereo track. Then with the cursor set at 0.0000, I shift the start time of the left channel to 0.0010. I made no other changes and left Audacity preset to Project rate 44100 Hz (as a matter of fact, I have never changed this rate).

The file now sounds much better, so I export it, leaving it in flac format.

As you can see the file has more than doubled in size. Could any of the changes I made have caused this increase?

======

General
Complete name : C:UsersUserDesktopBaby Elephant Walk after Audacity.flac
Format : FLAC
Format/Info : Free Lossless Audio Codec
File size : 30.8 MiB
Duration : 2mn 51s
Overall bit rate mode : Variable
Overall bit rate : 1 510 Kbps
Album : Bert Kaempfert And His Orchestra
Album/Performer : Bert Kaempfert / Bert Kaempfert
Track name : Baby Elephant Walk
Track name/Position : 14
Performer : Bert Kaempfert
Genre : Instrumental
Recorded date : 1978

Audio
Format : FLAC
Format/Info : Free Lossless Audio Codec
Duration : 2mn 51s
Bit rate mode : Variable
Bit rate : 1 510 Kbps
Channel(s) : 2 channels
Sampling rate : 44.1 KHz
Bit depth : 24 bits
Stream size : 30.8 MiB (100%)
Writing library : libFLAC 1.2.1 (UTC 2007-09-17)

Your original file was 16-bits. The edited file is 24-bits.

A 24-bit uncompressed file is 50% bigger than a 16-bit file. With lossless compression it’s not quite as predictable.

I shift the start time of the left channel to 0.0010.

That will cause comb filtering if you ever listen on a mono system.

Otherwise it’s a 1 millisecond delay in the left channel, which not that significant… You can get some comb filtering when the sound waves mix acoustically, but it’s no worse than having the left speaker about 1 foot farther from you than the right speaker, except with no reduction in volume. It’s probably comb filtering that changes when you move hour head that gives you a “spacious” sound that you find pleasing.

This will also make your FLAC bigger because FLAC takes advantage of the information that’s common in both channels and since you’ve shifted one channel there’s a lot less common information (at the byte or sample level).

When I make these ‘enhancement’ changes, I listen carefully several times during the process. Unless I am imagining things, the one millisecond time shift of one track does help me hear more sounds. At the very least, it causes me to notice sounds not noticed before. Unless those at the forum feel this is just my imagination, I will continue with this time shift.

About the bits. I have never made changes to depth or rate settings and never noticed these rates before or after enhancing a file. I’m guessing that most of the files I put into Audacity have a 16 bit depth and that due to a preset, come out as 24.

Where in Audacity would I find settings for bit depth? If it is preset to 24, I could set it to 16, unless you at the forum say this could cause a noticeable difference in the quality of the resulting file. Otherwise, I will just tolerate the increase in file size.

Where in Audacity would I find settings for bit depth?

After selecting FLAC, go to options. You can also choose “level” which affects the amount of compression. FLAC is always lossless, so it doesn’t affect quality. It just determines how “hard” the compression algorithm works.

If it is preset to 24, I could set it to 16, unless you at the forum say this could cause a noticeable difference in the quality of the resulting file.

You can’t increase quality by increasing the resolution… It’s sort-of like copying a VHS tape to Blu-Ray. And, 16-bits is already better than human hearing. However, the math involved in “processing” does “fill-in” the extra bits with information.

Unless I am imagining things, the one millisecond time shift of one track does help me hear more sounds.

I’m not sure… I’m NOT going to say it’s your imagination. You’ll certainly hear the difference in mono, but I’m not so sure about stereo. If you like the effect, it’s your music and you can do whatever you want! I do think this is contributing to file size. You can experiment with that if you wish.

If you want to take the time, there is software for doing a [u]blind ABX test[/u]. The way you do that is you listen to A and B and then listen to X, which is either A or B, but you don’t know, and it changes randomly. If you can identify X every time, you are really hearing a difference. If you get X right about half the time, that’s no better than guessing and you can’t reliably hear a difference between A & B.

An ABX test might be worthwhile if the time shifting is making your files bigger with no audible benefit.

A 10 or 20mS shift should be more audible and it will tend to “widen” the stereo. Sometimes a time-shift is used to get pseudo-stereo from mono source (a 2-channel recording with identical left & right channels). But again, this fouls-up mono playback. As the delay gets longer, the sound will seem like it’s coming from the speaker with no delay. By the time you get up to around 50mS, you’ll hear an echo.

It will be the change from 16 bit to 24 bit that makes most of the difference to the file size.
After selecting “Flac” in the Export dialog window (http://manual.audacityteam.org/o/man/file_export_dialog.html) click on the “Options” button and set Flac export as 16 bit.

P.S.

Speaking of ABX… You might try an ABX test between FLAC and a high-quality MP3. You might be surprised how hard it is to hear the difference! We’ve all heard low-quality MP3s, but often “night and day” differences and “horrible” MP3 sound quality disappear in blind listening tests with good quality MP3s. :wink: You can start with 320kbps or V0 which are the “best” MP3 settings.

But if you do switch to MP3, I’d recommend keeping your original FLAC archive.

MP3 doesn’t make very good archive because all you can do with the work is listen to it. If you try to make lower quality MP3s for your Portable Listening Device, the quality of the sound will be the combination damage of the two MP3s. That’s why even though additional storage medium needed, uncompressed WAV or equivalent archive is highly recommended. You can make those into anything.

Koz

I’m going to try the ABX test. It sounds interesting. I seriously doubt though that I would ever switch to an all mp3 library.I have listened to flac that was worse than mp3, about equal to mp3, better than mp3, and far superior to mp3. (I’m guessing this is due to how the file has been manipulated before I get it. I also feel that when I get a bad flac file it is usually due to over compression. Compression is one tool I use cautiously and sparingly).

With my library approaching 100,000 files, I firmly believe that as a rule, though not always, flac sounds much better than mp3.

I have learned a lot from this session. I’m going to keep my time shift technique, but make sure the files are rendered at 16 bits. This may keep the file size at a reasonable level.

One question before we close this thread: DVDdoug states that 16 bits is already better than human hearing. He’s the pro and I’m sure he’s correct. But, since 16 bits is already better than human hearing why are we offered 24 bits? It seems that would be jacking up the file size without any benefit.

Thanks for all the information

Doug

24 bit and higher have benefits during audio production. They provide more headroom in the digital domain, and greater precision when processing. Applying multiple processes to audio in 16 bit will have a cumulative effect - the quality will gradually degrade. 32 bit float format (used internally in Audacity) is extremely precise, so even with lots of processing there will be virtually no degradation due to rounding errors while in 32 bit float format.