Hi,
I want to publish some podcasts in internet and I’m testing what settings could be ok in order to have quality but not have a file of big size, in order to save band width.
I’ve done some tests with stereo and mono and also changing the fecuency (Hz) and I got surprised with two things. I think I’ve made something wrong.
I’ve done this:
I’ve recorded in stereo the same podcast six times, changing the project frecuency before recording. From 44100 to 8000 Hz
I’ve exported to mp3 the stereo version and also a mono version with “Split stereo track to mono”, to obtain a mono track and removing one of the two tracks obtained.
I thought that decreasing the Hz, the size of the file would also decrease. And also it would decrease after changing from stereo to mono.
Well, the two surprises I got are:
There is only a big difference of file size in the 11025 Hz y 8000 versions. The other ones are more or less of the same size and the difference can be because the time of the recording is not the same. And also the volume of my voice.
the file size is exactly the same between the stereo and the mono version, and I don’t understand why.
So I’ve two questions:
What am I doing wrong in my tests?
What settings do you recommend for a podcats for internet, with a good relation between quality and file size?
When using “CBR” (constant bit rate), the “bit rate” (kbps) that you select in the MP3 export settings is the bit rate that you get.
If you select 128 kbps, then there will be 128 thousand data bits for each second of audio, regardless of whether the track is mono or stereo. If the track is stereo then the MP3 will share the 128,000 bits between the left and right channels. If the track is mono the MP3 will use all 128,000 bits for the one mono channel. This means that for a given bit rate you will get better sound quality if the track is mono than if the track is stereo (except that it is mono ).
When using “VBR” (variable bit rate) the situation is different, For VBR (and for the “Presets”, which are VBR presets). the bit rate is calculated to achieve the “quality” level that you set. Mono tracks can achieve the same quality with less bits than stereo tracks, so when using VBR, mono tracks will be smaller than stereo tracks.
In the early days of MP3, some players had difficulty playing VBR MP3s. These days all MP3 players should be able to handle VBR and VBR will generally give better sound quality for a given file size than CBR. The only downside to VBR is that some players will report the duration incorrectly. If it is important that the duration is shown correctly in all players, use CBR, otherwise use VBR.
When converting, don’t manually convert the sample rate - leave that to LAME. LAME will tell you if it needs to change the sample rate. All you need to be concerned about is the final file size/bit rate and quality. Leave LAME to make the other decisions, it has been thoroughly optimised to make good choices
Usually a better way to obtain the mono track is to use “Tracks > Stereo Track to Mono”. This “mixes” the two channels into a single mono track.
For stereo music it is generally accepted that the minimum bit rate for reasonable sound quality is 128 kbps.
For mono music you can go down to 64 kbps.
If you use VBR settings for music, try “5 - 110 to 150 kbps”
For mono speech you can go a lot lower. The final decision about how low you can go is down to your judgement about what is “good enough”, but I would suggest that 32 kbps CBR should be good enough for speech. If you don’t find that is good enough, try “VBR: 9 - 45 to 85 kbps VBR (Smaller files)”.
Hi Steve,
thank you for you answer. Really usefull to me. It’s a speech, so I’ll try with 32 kbps CBR for example.
What about the frecuency (Hz)? What do you recommend?
Nobody understands compression. Compressors like MP3 strive to preserve the quality as much as possible of the original performance. If you pre-damage the original performance, then MP3 will try to carefully preserve the damage. It’s always best to start with the highest possible quality performance at the start of a compression job. That will result in the highest possible quality MP3 (or whatever compressor you’re using).
This, of course, kills people trying to make an MP3 from another MP3. The damage adds up.
Original quality doesn’t affect the MP3 file size at all. That’s a function of the creation of the MP3, the quality settings and the CBR/VBR thing. The only single show characteristic that will greatly affect the file size is Mono/Stereo.
If you’re interested in the smallest possible posted file, mono is without question the way to go. Many people can’t tell they’re listening to a mono or stereo file. Mono 32 and Stereo 64 are about the same sound quality with significantly different file sizes.
Another note. 32 is widely regarded as the lowest you can go for a performance that most people will accept. Below that and the show turns into bubbly, gargling, honky, grade-school trash. With the increase in quality of internet connections, those sorts of shows are declining in popularity because they can be hard to understand and enjoy. Audacity default is 128 and the iTunes best standard quality (other than uncompressed) is Stereo 256.
I listen to a podcast where the show is in reasonable quality, but the “commercial” at the beginning and end are highly compressed and sound like they’re talking through a long cardboard tube. Annoying.
As Koz says, the recording (before converting to MP3) should be a good clean recording. I’d recommend recording with a sample rate of 44100 Hz and 32 bit float. Leave the format like that throughout the production process. Convert to MP3 at the end and keep a WAV format backup copy.
Hi,
I’ve been doing tests and, as you said, for speech is perfect doing this:
Record stereo 44100 Hz 32 bits
“Tracks > Stereo Track to Mono”
Export mp3 at 32 kbps
I don’t notice any difference of quality between 128 kbps and 32 kbps. With 32kbps, between the stereo and the mono version I notice differences. The mono version is better.
).
