A Romantic Fangirl

How to recover a creashed Audacity recording

Sun, 04 Feb 2024 23:22:24 GMT

I recently had Audacity crash just as I stopped a recording and was about to save the project.
It usually recovers what you were working on when you restart it, so I wasn't particularly worried... Until I was greeted with a text box that said that it recovered to the last snapshot of the project and behind it, an empty project.

The best piece of advice for this situation I could find online was to find the temp file for the project and save it somewhere else as back-up. Nothing else I could find (change the extension name from aup3unsaved to aup3 and do the same for the other additional files - by cutting out the unsaved part of the extension - and open it directly in Audacity, or importing the temp file - which registers as an incompatible format), so I saved the temp files and decided to deal with it later.

After stepping away for a few days and spending a few hours playing around with various settings, I managed to recover a slightly damaged version of the recording (it's ok, the damage is mostly annoying)

problem: recording crashes before you can save or export it and when you try to recover it on opening audacity, you're greeted with an empty project
what to do:

Don't exit Audacity!!! you might lose the temp file
Set Windows Explorer to show hidden files
Go to C:\Users\[your user name]\AppData\Local\Audacity\SessionData and copy the three temp files for the most recent session to some place else (preferably, zip them together to make sure they won't go anywhere)
Open Audacity
File > Import > Raw Data and import the unsaved project from where you put your copy of it
You might need to fiddle with the settings on import (for some reason, this was saved at 22050Hz and you might need to change Byte order to Default endianness. If you get the rate wrong, you can change that in the project, afterwards)
You're probably going to get one huge file of everything that's been recorded that day, with gaps and glitches between sections of recording
Split stereo to mono and get rid of the channel with the egregious glitches
Export the file in a lossless format
You can deal with editing it later. At least you got your recording back

This is what the recovered audio looked like, before I did anything to it. (Ignore the 32-bit float part. This was an earlier recovery, before I remembered to change it to 16-bit and figured out the ratio)

comments

Behind the scenes: It's like Photoshop, except not really

Wed, 08 May 2019 21:28:57 GMT

I'm not sure part 1 of how many this is going to be, but I'll get around to doing a proper post about my editing process at some point.

You know those before and after Photoshop pics? Well, this is something like that, except for podfic. We were having a conversation in the server last night about playing around with volume and I shared a screenshot of a raw recording along with the finished podfic for comparison and then I went through a few more projects, because it looked interesting.

I thought I'd share a wider selection here for anyone curious.

I couldn't go back further than 2017, because any unedited files I have from before then are still unedited.

Here's a short one for a little breakdown

I tend to record each page/screen twice and mark the end with a snap. I count those as takes and any flubs and pickups in between those go unmarked. One page is about 4 minutes unedited and about 3 minutes when edited.
I use the second take and keep the first one as backup in case I mess up a word or there are any background noises I've missed when recording.
Most of the fics I record tend to be on the quiet side lately, so even when I boost the volume, they're not very loud. Still, I try to keep them from going over -3dB.

Here's a longer one (45min), recorded in two sittings.

This one's about the same length, except I had the "brilliant" idea to record it in one sitting. (It's Eat me, EL James! Yes, that's mostly one long sex scene)

This was a much longer one. This took three days to record and eight sittings. It was actually the third time I recorded this fic, but I never finished the first two attempts, because I kept having to change my recording setup.
For some reason, the project for the 6th part glitched and isn't working, but I had it saved in another folder and used that when editing.
Backup your raw recordings as WAV or other lossless formats before you start editing!!!

Here's a few more for the curious

The last one is one of the few times my raw recording was louder than the finished podfic.

comments

Remove mouth noises the hard way (Repair tool)

Wed, 23 Jan 2019 21:19:14 GMT

If you're particularly annoyed by mouth noises in your recording and you really, really want to get rid of them and you have enough patience to go through it by hand, the Repair effect is your best friend in Audacity. I suggest assigning it a keyboard shortcut or using Ctrl+R (repeat last effect used) a lot.
You find it in the Effects menu, under the Built-in section/the first part of the menu (depending on how you have it displayed). Myself, I've made Alt+Z my shortcut for it, because it's easier to remember along with Ctrl+Z for Undo.

I marked a few spots with noticeable mouth noise

When the noise is outside of a word, it's pretty simple to delete it - just do a fade in/out to whatever's next to it to make sure it doesn't sound jarring. Also, Repair helps here, too. Just apply it over the ends that are now joined.

When the noise happens on a letter, it gets a bit tricky and deleting tends to cause more problems.
You'll need to zoom in.

You can only apply the Repair effect over very short segments, so you'll have to go bit by bit until the wave is fixed. (Up to 128 samples or a few thousandths of a second - if you select too much, it will give you an error alert and do nothing, don't worry)

It doesn't always work perfectly, but it can get quite addictive (I have spent half an hour removing two seconds of phone interference from two minutes of recording - I was drunk on power and didn't care it would have been simpler to just rerecord).

You can also use it to turn an R into an L if you ever need that or, in a more practical application, to fix clipped sound waves.

Here's the portion of recording I used for the example. It's the same 17s first with mouth noise and then cleaned up

Also, pressing Alt while using the Draw Tool will give you a Brush Tool which you can also use to smooth out glitches and mouth noises in your recording, unfortunately, since the cursor doesn't appear in screenshots, I have no images for that.

comments

How to get rid of pesky sibilants without a de-esser [podfic tips]

Sat, 22 Dec 2018 17:06:33 GMT

Quick post, bc it's come up and it's something that took me a few weeks to figure out: How to manage sibilants without a de-esser plug-in in Audacity.

The short answer: use Plot Spectrum to find the loudest frequency and then tone it down using Notch Filter. Set Q to around 5-7 for best results.
Slightly longer answer with pictures:

Select the portion of audio that you want to de-ess. Go to Analyze and click on Plot Spectrum (I suggest adding a shortcut for this function if you need it often - which is how I managed to completely forget what it was called or where exactly in the menu to find it before I made these screencaps)

It will give you something like this, except the peaks should stand out more

Hover your mouse over it and it will show you the peak frequency

then use that in Notch Filter (love of my editing life!)

If you use 9 for Qq, it will cut out that frequency drastically. Use 5-7 to tone it down, 2-4 if you need to tone down the adjacent frequency bands.
It also works on lower frequencies, if you have a weird rumble that you can't remove with noise reduction or it's higher than 100Hz and a high pass filter would affect your sound quality too much

comments

Simplest Robot Voice Tutorial

Mon, 10 Dec 2018 11:40:33 GMT

xposting from Tumblr where I completely lost this post and reena had to find it for me :|

Right-click > save as, or left-click to play: sample

@reena-jenkins asked how to make a robot voice (a la Jarvis) in Audacity, so here we go :) I hope that’s close enough. I only used this once and the tutorial I learned it from was for Adobe Audition (surprisingly, the effects there don’t match)

I tried to take as many screencaps as possible, so anyone could follow.

You take your recording and make two more copies of it on separate tracks (just Ctrl+C/Ctrl+P twice in the empty area)

Then select each copy and go into the Effects menu to Change Speed. Increase the speed on one of them and decrease the speed on the other (I went with 1/-1 to keep it comprehensible. It gets distracting if you use more than 3/-3) Play around with the speed ratios until you get the type of robot voice you’re after. (-3/2 sounds pretty cool, but nothing like JARVIS)

Then you select all three tracks (Ctrl+A) and go into the Tracks menu. Click Mix and Render

And, voila!

comments