Hi Guys,

I'm working on some sounds for a website, and space is at a premium. The company have told me they usually compress stuff down to around 16kbps to enable the sounds to stream efficiently.

Are there any tricks I can take advantage of to keep the sound as high quality as possible at a bitrate that low? I remember reading in an iPhone dev article that you can pretty safely cut anything below 100Hz and above 18kHz since the player is likely using cheap headphones and the difference in sound will be negligible.

Anyone dealt with this kinda thing before?

Cheers Joe

4 Answers 4


Thought I'd chime in too. mono 48kbit mp3 does fine for sounds in full spectrum of up to 14kHz. MP3 and similar compression schemes work on the spectrum so the more you can chop off the more effective your compression.

That doesn't necessarily limit you to low frequencies. For some sounds you may have success with cutting the bottom end. Also if you're allowed to layer sounds that's more than doubling the bit-rate in question.

Certainly for anything speech, aggressive EQ prior to compression will give you more than it will take away.

  • Is there any information about this aggressive EQ you mention?
    – trlkly
    Oct 25, 2014 at 1:16

That's a huge compression ratio, man, and I don't envy your position. The only thing I can suggest is to work in as a high a sample rate as you can prior to encoding the compressed format. Compression codecs like mp3/aac/etc. Turn out better results with higher quality raw materials.

If you're recording the files yourself, go for 24 bit 192 khz audio files. If you're working with stock sounds, there's really very little point to up-sampling them, and you're going to be stuck with 44.1/48/whatever rate for the encode.

I don't envy you working with that delivery format. ;)


It is my understanding that evolution (LC-AAC, HE-AAC, HE-AACv2 in order of increasing compression ratio) of AAC provides improved performance over mp3 at the same bitrates, starting at 24kbps. For this purpose it uses tricks like Spectral Band Replication or "let´s keep the lower band up to 7kHz and add some info so the decoder will know what harmonics to regenerate" and Parametric Stereo or "let´s do it mono and encode which freqs go to left and right".

It seems that Dolby licences the same HE-AACv2 as Dolby Pulse.

Anyway, altough I haven´t done critical listening between AAC and mp3 I´m sure the first one would be the way to go if you want the best quality at lower rates.


I think it would be wise to check what codec will be used exactly, and do regular conversions to that format to check if what you are doing is translating to the low quality audio at all.

I recently did a project where the output was at 22 KHz mono, which already threw away most of the details and finesse that I put into the sounds.

16 kbps is terribly low, with any codec. Depending on which codec is used, it will sound different. So try it out (mp3 at 16 kbps barely produces intelligible speech, even though the human mind is super sensitive to speech).

But generally speaking, at such low quality: forget everything above 10 Khz or even lower, forget subtle dynamics or in fact any details, even in the time domain.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.