Many times I make a recording and decide to save it in MP3 format. Exporting as an MP3 takes many times longer than saving as a WAV or other less compressed format. I have often encountered the need to start a new recording before the previous EXPORT/SAVE operation has finished because it takes so long.
Many programs that can run for a very long time such as WinRAR, WinZIP and others, have a PAUSE function to suspend execution so that the PC can be used for other tasks without these programs sucking up a lot of CPU and disk I/O bandwidth. Indeed, saving a recording as an MP3 also seems to suck up as much CPU BW as it can get.
But BW is not the concern here. It’s annoying that another recording session can not be started without canceling a SAVE that’s often already an hour or more done, only has a little time left to go, and will have to be restarted from scratch. Being able to suspend (pause) an Export (SAVE) to start another recording session in a separate instance of Audacity would be a huge improvement IMO.
This does not directly answer your enquiry, but it’s a tip that you may find useful.
If you use “Save Project As” before you start a recording, then you will find that “Save” after completing the recording is extremely fast.
Thanks. One question is: in what format is the project saved? I record in 32 bit floating point, but when I save as a WAV (the least compressed format I use), it’s reducing the format to signed 16 bit. Presumably as a result of some roundings that take place during reduction from 32 to 16 bits, if a recording in memory that was saved as a signed 16-bit WAV is opened and saved again in the same format to a new file, there are many differences between the two files. I don’t know whether there is any theoretical degradation if a recording saved as a project is reopened and then saved as a 16-bit WAV vs saving the recording directly as a WAV.
But more to the point, I often have many instances of Audacity open that have not been saved and can’t always anticipate when I will need to perform an additional recording. Eventually each of these open and unsaved recordings needs to be saved and often, while that is being done, an unexpected need to record arises and the save must be canceled even though it’s already been running 60-90 minutes.
If the project “Quality” is set to 32-bit float (as it is by default) then the audio data is saved in 32-bit float format.
An Audacity Project is not a file - it consists of multiple parts: a “project file” and “audio data”.
See here for more detail: http://manual.audacityteam.org/man/Audacity_Projects
When exporting to a 16 bit WAV file, by default Audacity will perform proper 32-bit (float) to 16 bit (signed integer) conversion (in the same way that professional recording studios convert 32-bit master recordings to CD masters).
Audacity does not (normally) record to memory. The recorded data is written to disk.
If the project has not been previously saved it will be written to a temporary folder.
If the project has been saved it will be written to the project “_data” folder.
When a previously unsaved project is saved, the audio data is copied from the temporary folder into the project _data folder and this is why large projects take a while to save. If the project has already been saved previously then the audio data is already in the _data folder which is why saving is so much faster.
I hope that your computer is unconditionally stable.
Audio recording is quite demanding and there will be an awful lot of data hanging around. I would not normally recommend having multiple unsaved projects open at the same time as most people have computers that are not terribly stable and a crash would be disaster.
Try the “save it first” tip, it should make saving projects far less arduous for both you and your computer.
The original poster was talking about “saving” in MP3 format; i.e. he/she was doing a File > Export operation. I’m curious to know how big these recordings are that can take 60-90 minutes to export as MP3. The impression I have formed when exporting MP3 files is that they take about 20-25% of the actual recording duration to complete the export (i.e. a 10 minute recording would take about 2 minutes to export). If this user has several ultra-large Audacity projects open at once in his/her system, I’d expect that the system would be spending all its time paging in and out from backing store and having little time left to do the real work!
Based on the Windows file Creation and Modification time stamps and one example that is typical, a 2 hour recording took 31 minutes to save as an MP3. So that’s 25%, as you said.
While saving as an MP3 the CPU is nearly saturated. Judging from Windows’ Task Manager’s display of CPU real time usage numerical display is typically in the 90% range. It’s a 2 GHz P4. Saving as WAV is of course much much much faster.
One solution would be exporting as .wav or .flac, then doing the conversion to .mp3 with a separate program (e.g., SOX). The export will complete fairly quickly and the separate program won’t lock Audacity out of starting a new project. He didn’t say what his computer is, but if it’s multi-core, SOX runs in only one CPU (because it’s single-thread) so Audacity has the rest of the machine. Disc contention shouldn’t be a problem either - there’s enough arithmetic to generating mp3 to ensure long gaps between disc accesses by the second program, thus disc availability to Audacity.
Exporting as .flac takes longer than .wav, (although there’s a setting to minimize the time) but allows filling in the metadata, which the second program will (one would hope!) propagate to the mp3. I’m not sure whether .wav exports actually use the metadata. Is there another lossless export that’s quick and also saves the metadata?