how to reconstruct audio recording with only .au files

Hello everyone –

I need to reconstruct two audio interviews recorded by another person (using Audacity), who has been able to provide me only with the .au data files originally associated with the project.


She no longer has the “project folders”. :blush: Each interview is about one hour long, so I have dozens of .ay data files for each interview. :laughing:

How do I paste these together? Perhaps the question is how might I reconstruct the original two projects? Ideally I believe I’d like the two projects to be in mp3. When I open each data file in Audacity, I can hear the original recording --for 6 seconds.

I have AUDACITY 2.4.2. I am using a Mac Book Air with Catalina 10.15.5. I am a complete newbie at Audacity (* love * the name!!).

Many thanks for any advice!!

All best to everyone -


Does the other person have the AUP files?

If not it is just about impossible to reconstruct the recordings.

– Bill

They have not been edited or changed in any way, right? If you do that to an edited production, that’s the end of the world. The two interviews are in separate baskets?

There was a poster a while back that did manage to smash a pile of AU snippets back into a show.

I think it’s this one. I did save it.

This is not fun and some of it is by guess and by golly.

The Audacity Project format is being changed so it won’t do that any more.
No more thousands of little AU files. Yay!


Hello Audacity Forum, Bill and Koz !!!

Thank you so much for responding so quickly!

So to answer the questions:
I do have all the .au data files.

The two interview projects were sent to me via DropBox, in separate folders (baskets?).

Each interview folder contains 4 or 5 folders, within which there are dozens of .au files → presumably (hopefully!!!) in chronological order.

I do not believe these .au files were changed or renamed in any way.

I will now read the poster Koz sent (with a strong drink in hand) and try to learn how to smash the pile of snippets together.

I’ll keep you posted.

You have given me some hope. *** Thank you so much!!! ***


PS – I do have full written transcripts for each interview!!! So that should help when I’m trying to piece the snippets together.



So… Next time you’re going to save a WAV file (or AIFF) back-up file, right? :wink: I always recommend exporting to WAV immediately after recording, whether you make an Audacity Project or not.

And if it’s super-important, record with your phone (or a stand-alone solid state recorder) and the computer at the same time for safety/redundancy.

presumably (hopefully!!!) in chronological order.

The filenames are completely and intentionally scrambled. I don’t remember the exact reason they did that, but there was a reason. Older projects didn’t do that and while it made reconstructing a damaged Project enormously easier, it causes other very serious problems and accidents.

That’s why you have to do all the reconstruction with the file Time and Date stamps—and they’re not good enough. You noticed that the snippets are all 6 seconds long. Windows time stamp (I believe) takes you down to the minute, so you will have 10 sound files, randomly named, with that stamp.

It’s up to you, by ear, to shuffle the files until they make sense. The transcription will be an enormous help, as will noting the beginning and endings of segments on both sides. "The last segment ended in “eh-”, so the next snippet in the next segment should start with “-nter” for the word “enter.”

If it was a stereo show, there will be two very similar snippets; they alternate left and right. Wear headphones so you can tell if the show reverses left and right at odd times.

There’s 60 bunches of files per show and you have two shows.

This is why people opt to re-record the work rather than trying to jam all those back together.

I will now read the poster Koz sent (with a strong drink in hand) and try to learn how to smash the pile of snippets together.


There was a thought experiment a while ago which posited that in any one segment, the ends of the six second snippets should be close to each other and the tiny sections of the blue waves should meet or move in rational directions. It should be possible to get really close automatically and then only clean up the one or two segments that the automatic service just couldn’t figure out.

For example, these are likely to be a match…

Screen Shot 2020-11-11 at 15.52.32.png
…whereas these two are not…

Screen Shot 2020-11-11 at 15.50.19.png
As has been noted, there’s no shortage of people volunteering jobs for the programmers and developers to do.

It was determined that changing the Project Format so it never-ever did that again was far better use of the programmer’s time.


Another thought experiment is to take a big bag full of sentences and try to reconstruct a novel.

(1) It was a dark and stormy night.



Hello again Koz and company,

Thanks so much for the additional information and advice and pictures.

Yes, I discovered that the audio files are in no order. (AAAUUUGHHHH.)

I’ve never edited an audio file before, so my learning curve is essentially vertical.

I’ll try to tackle this over the week-end.

I haven’t yet reviewed Audacity’s instruction videos, but might you be able to point out where the most basic instructions are re how to add one sound file to another? (This may be so basic and obvious to others that it falls under the category “intuitive” → at which point I’d be in very big trouble.)

I did a test to put three .au files into the record page, and when I pressed the on button it sounded to me like I was listening to all three tapes at the same time.

The transcript will be essential; I see that I can search the transcript and locate where the audio file belongs.

Thanks so much Koz et al.



it sounded to me like I was listening to all three tapes at the same time.

You can’t stack them up and down. Audacity will play those at the same time. That’s intended (for example) to put your voice on track 1, background music on track 2 and maybe special effects (rain shower) on track 3. You can select each track and make them louder or quieter as needed for the story, move them sooner or later, etc.

You need to place the snippets one after the other left to right.

You are witnessing the yawning abyss that is my editing skill. We need to wait for someone who actually knows how to edit without fumbling and looking everything up.

There is an up side to this. You’ll go into this a rank amateur and, if you’re successful, leave a seasoned professional. This is a lot like how your first successful audiobook works.

I’m guessing if you get good at this, you can sell your skills.


I can think of one way that would work. Open each similarly-dated sound file on its own track top to bottom. Use the Time Shift Tool (two sideways black arrows) to push each track left and right until the order makes sense when you play them. As you noticed Audacity will play everything unless you stop it with the SOLO and MUTE buttons on the left.

Also be clear that the snippets will flop rapidly Left and Right if it’s a stereo show.

Screen Shot 2020-11-13 at 1.01.01 PM.png
There will be a little yellow line when you get the alignment between two snippets perfect.

Since they’re random, you might have this. That will play 1, 3, 2, 4.

Screen Shot 2020-11-13 at 1.03.04 PM.png
And stereo might look like this.

Screen Shot 2020-11-13 at 1.05.07 PM.png
You can convert each track pair to actual stereo with the menus on the left. Select the top of the two. Make Stereo Track.

Screen Shot 2020-11-13 at 1.13.24 PM.png

I can hear the actual editors laughing at me for doing it this way.


Koz, you have become a brilliant shining star in my firmament of heroes. Thank you for all your time and help --and especially for imagining me as the future go-to person for miniature-file-stitching!!!
More anon -

calcagno, now that you have become an expert at editing audio in Audacity :wink: , you might also check out the following link:

Skip steps 1-7 and go directly to the time-saving alternative.

This aligns the selected tracks one after the other so they follow the Timeline.

How does it “know” which snippets come first? Does it do it top to bottom?

I don’t see how that would work with ten randomly named files with the same time and date. If you’re rescuing a dialog or conversation, you have to get the snippets close enough to each other to tell what the conversation is by playing it.

What does it do to this segment?

It turns out those snippets theatrically follow each other.

Or is the solution that you just have to get close and the tool will produce the final critical alignment?

Still, nobody knows what the order is going to be and how many files there are because the show may or may not be stereo.


At the risk of not following the show format, I think I would identify all the Lefts and Rights in one segment and rather than trying to straightened them out, Mix them and cut the show Mono. Tracks > Mix > Mix Stereo down to Mono.

Then the alignments reduces down to First Sentence > Second Sentence > Third…etc.

Do Not do any of these experiments on the only copy of the show. Make safety copies on thumb drives, external hard drives or cloud storage.

Think of how much fun this would be if you got half-way through successful patching and your computer’s eyes rolled up and it started blowing spit bubbles.

Windows has encountered a fatal error and will close.

As you progress, you might also consider Saving the work so far as a LossLess Project (Save-As) with a unique filename.


I would probably try this mono thing on a couple of segments before you commit the show. It’s possible the mono mix-downs will not marry to each other end to end without a tiny pop or tick.