Making educational screencasts with Adobe Presenter I record voice-over audio in Audacity with a Blue Yeti microphone. This mic has a sampling frequency of 48,000 Hz and a sampling depth of 16 bits. Now does this mean (as I suspect) that the optimal sampling depth setting in Audacity is also 16 bits, or would a higher setting (such as the default 32 bits float) still bring any benefits.
Audacity 2.0.6 in Windows 8.1 on a PC with an i7 processor and 12 GB RAM. .exe installer.
The optimal “sample format” (“bit-depth”) for Audacity is 32-bit float (always, regardless of the microphone / sound card being used)
32-bit float provides much better processing precision, and because Audacity uses 32-bit float format internally, it avoids unnecessary losses due to converting between formats. Also, modern processors (just about all processors of the last 15 years) handle floating point calculations more efficiently than integer calculations.
48000 as a fuzzy generality is a video sound sample rate. 44100 is the Audio CD sample rate. The earth will not stop going around if you cross them. Most video editors will accept 44100 without even blinking.
However the show will be missing one tiny conversion if you get them right.
Most of my sound stuff was shot at 48000 Stereo because I knew the work was going straight to Video Editorial.
Audacity will volunteer to convert the work to 16-bit when you export. You should probably take them up on that.
Koz, I want to make sure I understand this. With your last statement is it safe to assume when recording for VO I should have my bit rate at 32 but when I export to a wav to send to a client I need to change it to 16-bit? This is all foreign to me so trying to make sure I get it right. Do I record at 16 bit and export the wav keeping it at 16 bit or record at 32 and export the wav keeping it at 32 or change it to 16?
I doubt the microphone will let you connect at anything but 16-bit. The up conversion from 16-bit to 32-floating in Audacity is perfect. It’s coming back down you have to be careful.
32-bit is different from 32-floating. As Steve above, Audacity works internally at 32-bit floating unless you force it not to. 32-floating doesn’t overload and it allows you to do effects, filters, modifications, and corrections without worrying about clipping, peaking, and overload distortion. The only down side to that is the need to convert the show to either 44100, 16-bit or 48000, 16-bit for client delivery.
If you’re live recording, it’s totally possible to overload the microphone, interface, mixer, or sound card before Audacity. Running red clip or peak lights is never a good idea.
You should find out what the client wants. The most common uncompressed sound format is 44100, 16-bit, WAV and that can be a good starting place if you have no idea what the client wants. We had a production recording engineer do a voice recording and he didn’t ask. He delivered in a super-duper high quality format and a simple MP3. Nobody could open the good quality sound file, so we got forced into using the MP3.
MP3 should not be used for production. It’s an end-product format you use to put music files on your iPod for running on the beach. MP3 gets its small convenient sound files by leaving out certain sounds and cleverly hiding sound distortion.
Further Audacity oddity. Audacity doesn’t save sound files. It saves Projects which are unique, special-purpose products that can save edit timelines and display formats in addition to the actual sounds. Projects are a little odd, only open in Audacity, and they will not save UNDO.
To get a ‘real’ sound file you have to File > Export one. I export all raw voice work as WAV before I do anything to it. Mistakes, fluffs, repeats and all. There should not be any reshoots if Audacity or the computer stops working during an edit.
That and I occasionally get calls from Producers wanting good quality, unedited copies of sound shoots.
Thanks! So to clarify, if a custom wants 16 bit I should record in 32-bit floating and then convert to 16 bit after it is recorded/edited? Or if I know the customer specified 16 bit / 44,100 I should just record from the beginning in 16 bit?
I don’t know how you see this on Windows, but there is a utility on a Mac called Audio MIDI Setup that allows me to see the digital standard that the Mac is using to connect to my USB thing. I don’t have a USB microphone, but this is my USB stereo adapter.
Note the standard is Stereo, 16-bit, 44100.
At the same time, Audacity > Preferences > Quality set to 32-float. That’s the point where Audacity converts from 16-bit to 32-floating. That’s the default and you probably shouldn’t change it.
Hi. Long story but I have a mobile recording booth and 4 acoustic panels. I’m getting ready to go to bed so didn’t want to fuss with setting it up so just put an acoustic panel in front and one behind and stacked one as a cloud over head and recorded real quick as prescribed. It is attached. No editing and I think I followed your instructions exactly. I recorded in 32 bit floating/44100 and exported at 16 bit PCM. I didn’t put any effort into the acting and basically just read your script and cut the file at 10 seconds.
Interesting. I used the raw audio I sent to you and just rehearsed my normal de-ess, de-click, EQ, compress, normalize and was listening to it compared to my raw. I imported the saved file I sent you into Audacity. When I recorded your audio I kept it at 32 bit float. I absolutely changed to 16 bit PCM when I exported the wav file. When I import that same wav file I exported back into Audacity it shows the file is at 32 bit float. How is that possible if I changed it to 16 bit PCM when I exported it?
By default, Audacity converts files to “32-bit float PCM” on import. This conversion is 100% lossless and it minimises quality losses when processing the audio. Processing as 16-bit (integer) would lose a little bit of sound quality with every effect, but 32-bit float doesn’t.