roll and punch in audacity

I record audiobooks using Audacity, and I’d like to experiment with roll and punch recording. I’m interested if anyone has ideas how I might do this in Audacity.

During the recording it is inevitable that a narrator will make mistakes, or decide to re-read a phrase to adjust the emphasis. If you do a straight read through you have a task afterwards to edit out the flubs. This secondary editing task can more than double the time taken to produce final audio.

Currently I mark errors with a clicker, so that you can visually identify where in an hour long file to make the edits. This is very quick, and I can code what sort of flub I have made - one click for a flub and repeat, two clicks for background noise etc… However, other narrators tell me that using roll and punch editing you can save some time by doing most of the editing during the recording session.

For those who don’t know what this is here is my understanding of what you would do:

  1. Flub and stop recording
  2. Mark the punch-in point (at a silence - usually the start of a sentence or phrase to help with consistancy of voice)
  3. Hit a key to start. Computer plays the audio that comes just before the punch point (set by pre-roll variable) and then starts recording on the same track at the punch point.

Now I know from reading around that Audacity does not implement punch-in recording in that way, preferring to spawn a second track (I can see the advantages of this - don’t get me wrong), but can any of you brains think of a work-around to get this type of recording on Audacity. Bear in mind that I may be making as many as 60+ edits on an hour long file (think of how many tracks that would be - and how long it would take to consolidate them into one audio track), and recording ten or twelve hours of audio for a medium length novel. Even numbers of keystrokes becomes significant with longform voicework solutions!


The old Audacity used to have problems, but not 1.3. Audacity 1.3 has Append Record. Stop when you make a fluff. Delete everything after the fluff. Press Append Record (Shift-R), and Audacity will pick up at that point without a second track.

You know Audacity will allow you to post Labels on the fly? Make a fluff, press Label (Control-B or Control-M, I forget). It’s in the Tracks menu. This will let you keep going through the whole script and then go back later and find each label and fix the dialog.


That’s really interesting. That would get around the multiple track problems completely. However I still want to be able to replay the preceeding 5 or 10 seconds and pick up the reading at the same pace. Sometimes the punch-in would be within a sentence.

It sound slike this is really close to working. Perhaps I could write a script to do this?

If I put the cursor at the punch-in point I could delete from the cursor point to the end of the file, back the cursor up by the pre-roll value, play the pre-roll portion of audio, and then append record from the end of the file.

Can I assign scripts to hot keys?

There are hundreds of Audacity users at LibriVox who could use this if I can get it right!


“Punch-in/Punch-out” recording was a crude form of editing that was invented back in the days of tape recording. The problem with this type of editing, and why it is never used in professional recording these days, is that it is virtually impossible to match up the waveform perfectly at the start and end of the punched section.
Common techniques that are frequently used in preference to punch-in/out are:

  1. Mark the “flubs” (can be done during the recording, or after) but don’t stop, just carry on.
  2. Mark a “flub” at the time that it happens, and immediately re-read the sentence or paragraph without stopping or pausing the recording.
  3. Mark a “flub” at the time that it happens and pause the recording. If it’s a difficult section to read, read it through a couple of times without recording, then un-pause and read the entire sentence or paragraph.

Necessary remedial action is then done after the initial recording session is finished.

Different approaches suit different people, but punch-in/out is probably the least likely to produce a convincing professional sounding result.

Sometimes the punch-in would be within a sentence.

Replace the sentence. Sophisticated non-linear editing was developed to get around problems like this. I remember punch-in. It was a nightmare because you never knew if it worked or not until you got the rest of the show down – and about a third of them failed.

Play the work until the end of the next working sentence, delete everything else then Shift-R Append Record.

It’s highly recommended to use the editing features. Plant a label when you make a fluff, repeat the dialog and keep right on going. Use the label later to go back and accurately place the correction.

“…was expected to repair the problem caused by freeb !@#$% [Label]…repair the problem caused by the freedom of the actuator in the propulsion system.”

I do multiple edits like this every week and navigation, zooming, deletion, and playback are very fast and accurate with keyboard shortcuts. Given the work was captured in good order (no DC level or other problems), you can produce flawless, cadence-accurate performances using this technique. And UNDO them if you feel like it.

Since I know this is going to come up. DO NOT edit and record the corrections on the original capture files and Do Not edit MP3.


I suppose you could use the overdubbing technique – play the drum track while you’re recording the guitar. That will give you perfect cadence matching, but that gives you two tracks again.


I take what you are saying, but remember we are talking solo spoken voice here. There are plenty of silences to hide the edit itself, you make all edits at phrase ends or even better at the start of a sentence. A slight change of pace or vocal tone is much less noticeable when you start a new idea, so edits between sentences and paragraphs are best. It can be technically hard to match drop-ins mid phrase as you say, but most narrators won’t drop-in mid phrase because it is so hard to match exact vocal tone and stress mid phrase and so easy to simply redo the whole sentence.

Yes. The commonly used amateur technique is to use technique 2 or 3. Technique 1 above means you have to return and re-record a drop in, which always takes longer. It is always harder to match exact vocal tone and pace at a later date. Much of the key to speeding up narration is to do as much as you can in the one session.

The “roll and punch” technique is the one used by most professional narrators (as far as I can tell). You aren’t punching in to the middle of an already recorded track, you are appending the recording on tp the end of a partially completed track. The special bit is just to allow you to hear the last bit of recorded audio so you can pick up the pace and tone and then start the recording automatically.

I want to be able to do this with the fewest number of keystrokes. I think it is hard for anyone who has not done this sort of work to understand howmuch time can be taken up by slight increases in number of keystrokes. What I would like to set up is

  1. one keystroke to pause recording
  2. one mouse click to reposition the cursor at the start of the last sentence
  3. one keystroke to delete from cursor to the end of the track, playbck 5 seconds of audio preceeding the cursor psition, start append recording.

I’d agree entirely. It is hopeless when you want to drop in corrections when you return to a recording with proof-listener corrections. To do these edits I record a second track and insert it into the main audio afterwards. I’m not talking about punch-in/punch-out editing, but a special case of record-append. The reason it is called “roll and punch” is that narrators have adapted bits of software (e.g. ProTools) that were originally intended for use in punch-in/punch-out editing.

hope this clears things up

I have found a video clip of someone using this technique. I’m nervous of posting this clip because:

  1. It’s from Audible. An Amazon company. Nuff said.
  2. The audio editing is awful! On a video promoting audio editing!

Click on the Recording and Editing button, but just watch fom 5min to 8min. It’s not that easy to see what the narrator is doing, and the guy explaining does not really go into why this technique is useful… But this is what I understand many pro narrators do. I have been narrating for years marking flubs and editing later. I’d just like to have a go at this and see if it will save any time.


I can get you close to that :stuck_out_tongue:

Everything is set up and you are happily recording, then there’s a “flub”.

  1. Press Space. Recording stops.
  2. Click on the track at the beginning of that last sentence (you can probably pick it out just by looking at the waveform), and drag to the right beyond the end of the track. The entire “flub” sentence is highlighted.
  3. Press the Del key. The flub sentence is deleted.
  4. Click on the track about 5 seconds before the end of the the track.
  5. Press “R”. A new track is created and Audacity is recording to the new track.

At the end you will have multiple tracks, but that’s OK. When you Export they will be mixed together, or if you want to do more editing you may wish to mix all of these tracks down to just one track - to do that, “Ctrl+A” (select All) then select “Mix and Render” from the Tracks menu.

We are assuming here that background noise is negligible, but if the background noise is at all audible, you may want to trim off the overlapping “silences” between takes before you mix down. The quick way to do that is to click on the lower track to the left of where it starts, then drag to the right until directly below the end of the track above (you will see a yellow vertical line when you are exactly lined up with the end of the track above). Then use “Ctrl+Alt+K” to “Split delete” the unwanted overlap.

We are also assuming that your drop-ins are at exactly the right moment for a smooth and natural flow. If they’re not, switch to the “Time Shift” tool (F5 key) and adjust the position left/right as necessary.

The process is a bit different to using ProTools, but having used both I find Audacity faster.

Steve –

The sequence you describe works just fine for me. The studio where I sometimes record has custom software to handle Punch-In, so I’m used to doing this from the engineer’s booth.

I was looking for a way to replicate this method in Audacity, and I’m not sure I would have discovered this particular sequence – and boiled it down to the simple steps you provided – without a lot of fooling around.

Thank you!

Glad you found it helpful Bob. I must get round to writing up some of these techniques on the Audacity wiki.

[Moderator note: more information in this topic:]

I was researching how to do non-destructive Punch and Roll in Audacity and eventually realized that I could simulate the feature using AppleScript on Mac OSX.
My method relied on a lot of existing Audacity shortcuts run together like a macro would be.
Hopefully, someone else will find this helpful.

Hi, I was asked to make Yoga audiobooks in exchange to become a yoga instructor. I was given a studio in Istanbul but as I changed my location now, I have to do all the recording from scratch with the new Zoom H6 I bought to record this audiobook. I have been looking for the phrase: punch n roll recording. I discovered the phrase on this video:
It explains how to do punch n roll recording using a free software called Studio One Free.
Did you find anything more useful using Audacity?
I am very new into making an audiobook with my own voice so I will be greateful for any help. Thank youç

In response to an assertion that punch in is crude and never used by professionals, I submit this: On a DAW that supports it, I bet Clive/fizzcat could quickly master punching in at any pause before the flub. Workflow: Start recording voice. Make mistake. Stop recording. Roll back a few seconds, or maybe a sentence before the mistake. Playback. Punch in at a convenient pause before the mistake and continue recording voice until the next mistake happens. Repeat. It isn’t difficult. There is no faster way to edit. Also, unwanted verses or solos in stereo mix recordings can be removed by punching in on the beat as the offending part begins and punching out on the beat after the offending part ends. Then hit and close the gap. You could mark in and mark out to preview the edit and adjust points before committing. It helps if you can feel the beat of the music. Try it, and you might be amazed. -dBu

“Punch-in/Punch-out” recording was a crude form of editing that was invented back in the days of tape recording. The problem with this type of editing, and why it is never used in professional recording these days, is that it is virtually impossible to match up the waveform perfectly at the start and end of the punched section.

The punch-in accuracy will be off because of that “one computer late” thing. Plus control latency. Audacity doesn’t really “know” what you’re doing. That and Time Shift Tool can only be executed easily from the left and you have to remember to select all succeeding tracks as you go.

But, yes. That’s the closest technique I’ve heard to do that.

Does that delay go away if you do it in Overdub mode?