resampling in Audacity?

[brad.bartels]

I have a show I recorded at 24-bit / 48000 Hz. I pulled it into Audacity (ver 1.2.6) to split the tracks and (I thought) converted to 16-bit / 44100 Hz. I normally record that way for simplicity, but I wanted to try some different / higher rates. From what I've read, I should be able to set my Uncommpressed Export Format to 16-bit WAV to get it from 24-bit to 16-bit. That part seems to be fine. But for the sample frequency from 48000 Hz to 44100 Hz, my understanding is that I should change the project sample rate in the lower left-hand corner to 44100 and export the files? When I do that, I get empty audio (~ 3 to 4 sec at the end of a 4 to 5 minute track) at the end of each file. I also exported the same tracks with a default project rate of 48000 Hz. Those files do not have the silence at the end and both tracks (the 48000 Hz and the 44100 Hz) are the exact same filesize and length. Nothing seems to be missing from the audio in the 44100 Hz version, so my only conclusion is that Audacity is compressing the audio and speeding it up a little bit? Something is definitely amiss here, could definitely be user error, but does anybody have an idea what might be going on? I've listened to both versions and I can't swear that the 44100 Hz is sped up, but it would seem that would have to be the case. Knowing what I know, I can at least imagine that the 44100 Hz version is slightly faster.

If there is a different / better / correct way to resample from 48000 Hz to 44100 Hz with Audacity that somebody knows about? Any help would be appreciated. Alternatively, I could take the 48000 Hz files and convert them with something else if that is easier, I guess. What do other folks use to resample anything not at 44100 Hz (free, hopefully)? I think I've used Winamp in the past, but I've played around a little bit in there and I can't remember how I did it.

[flipp]

I haven't used Audacity in quite some time so can't help you with it but for a free, standalone sample rate converter I can recommend r8brain - http://www.voxengo.com/product/r8brain/

[brad.bartels]

I haven't used Audacity in quite some time so can't help you with it but for a free, standalone sample rate converter I can recommend r8brain - http://www.voxengo.com/product/r8brain/

Thanks - I saw that in some other threads and I just downloaded and installed it. I took one of the tracks that I had exported from Audacity at the 48000 Hz sample rate and converted it with the r8brain software. If everything was kosher, the length (in min:sec), should be the same as the original track, shouldn't it? I mean if the speed is correct? The file I got out of the r8brain software is smaller in size, but also 3 seconds shorter than the original track. Maybe I need to convert the raw file to 44100 Hz before I even import it into Audacity? I could do that, I'd just have to split everything into tracks again. Kind of a pain, but if that will work, I'd be glad to do it. The recording sounds very nice, I want to get it converted properly.

[page]

audacity has a bug when you resample during export.

To avoid it, resample the entire show as a single track, then chop it up during export (via labels, or however you prefer, I recommend using labels for ease though). The option is under the Tracks pulldown, around the 4th option.

Last, upgrade to 1.3.x, yes it's in "beta" but it whips the pants off of 1.2.6 and ever since about 1.3.6 I've found it more stable.

edit: also, dont forget to change your project sample rate, bottom left corner. This should match what you want the final rate to be even though you are resampling it via the menu noted above.

[doodee]

That's the answer. I was having an issue where a few seconds at the beginning and end of each track would repeat when I exported them. Finally figured out it only happened when I was downsampling and splitting tracks at the same time.

[brad.bartels]

Thanks doodee and page. Sounds like that is the problem, I'll give your solution a try - tomorrow, it's past my bedtime:). Much appreciate the help on this.

[willndmb]

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

[page]

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

Be careful about dither. Audacity dithers regardless of change in bitdepth (or direction of bitdepth) if you've set it to dither on export. So unless you're flipping that setting deep in the preferences menu each time, you're dithering twice at 16bit.

It's a handy program, and I can't find anything better at it's price point, but it has a ton of quirks like that.

[brad.bartels]

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

Be careful about dither. Audacity dithers regardless of change in bitdepth (or direction of bitdepth) if you've set it to dither on export. So unless you're flipping that setting deep in the preferences menu each time, you're dithering twice at 16bit.

It's a handy program, and I can't find anything better at it's price point, but it has a ton of quirks like that.

I had downloaded the Voxengo r8brain that flipp referenced at the beginning. When I processed it the first time in Audacity, I needed to boost one of the channels a little bit to even them out and then normalize / amplify it by like 4 or 5 dB if I remember right. I'm thinking the best approach would be to run my raw files (it was more than 2 Gb worth, so the recorder of course broke it into 2 files) through the r8brain software to resample it to 44.1 KHz but leave it 24-bit, then import that into Audacity and do my normalization / amplification and split tracks and then export as 16-bit .wav's. Does that sound like a reasonable approach? I was hoping to not have to retrack it, but it looks like I'm going to have to do that anyway, so it seems like it might be best to resample to 44.1 kHz before even importing into Audacity and then do all the editing in Audacity while it's still in 24-bit format? Really appreciate the help on this.

[page]

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

Be careful about dither. Audacity dithers regardless of change in bitdepth (or direction of bitdepth) if you've set it to dither on export. So unless you're flipping that setting deep in the preferences menu each time, you're dithering twice at 16bit.

It's a handy program, and I can't find anything better at it's price point, but it has a ton of quirks like that.

I had downloaded the Voxengo r8brain that flipp referenced at the beginning. When I processed it the first time in Audacity, I needed to boost one of the channels a little bit to even them out and then normalize / amplify it by like 4 or 5 dB if I remember right. I'm thinking the best approach would be to run my raw files (it was more than 2 Gb worth, so the recorder of course broke it into 2 files) through the r8brain software to resample it to 44.1 KHz but leave it 24-bit, then import that into Audacity and do my normalization / amplification and split tracks and then export as 16-bit .wav's. Does that sound like a reasonable approach? I was hoping to not have to retrack it, but it looks like I'm going to have to do that anyway, so it seems like it might be best to resample to 44.1 kHz before even importing into Audacity and then do all the editing in Audacity while it's still in 24-bit format? Really appreciate the help on this.

If you did the tracking via labels, you can export the labels as a text file and re-import them later if you decide to make changes.

Without doing Audacity (combine, edits, export as 32bit due to edits & other processing) > r8brain (resample, dither to 16bit) > Audacity (track) then what you're suggesting makes the most sense. I've heard it's best practice to do edits in the largest domain possible (whatever your raw file comes in as, in this case 24/48), and resample, then dither at the end (in that order), but I'm not sure how much it really matters in the grand scheme of things versus resampling first.

If you can run r8brain as a plugin in Audacity (like as a VST or AU plugin), then you could do the resample/dither last, set your project workflow in the bottom left corner as what you set it to in r8brain, and (if r8brain will do it as a plugin) turn dither off. I've used Ozone from Izotope that way in conjunction with Audacity.

The bummer about all of this is it's a bell curve. When it's done perfectly, it's great (you don't notice any problems). When it's done sort of haphazardly, you probably won't notice any problems, or at least probably won't attribute them to resampling/dithering (because it's under the noise floor), but when you do it poorly, its sort of noticable during critical listening.

[brad.bartels]

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

Be careful about dither. Audacity dithers regardless of change in bitdepth (or direction of bitdepth) if you've set it to dither on export. So unless you're flipping that setting deep in the preferences menu each time, you're dithering twice at 16bit.

It's a handy program, and I can't find anything better at it's price point, but it has a ton of quirks like that.

I had downloaded the Voxengo r8brain that flipp referenced at the beginning. When I processed it the first time in Audacity, I needed to boost one of the channels a little bit to even them out and then normalize / amplify it by like 4 or 5 dB if I remember right. I'm thinking the best approach would be to run my raw files (it was more than 2 Gb worth, so the recorder of course broke it into 2 files) through the r8brain software to resample it to 44.1 KHz but leave it 24-bit, then import that into Audacity and do my normalization / amplification and split tracks and then export as 16-bit .wav's. Does that sound like a reasonable approach? I was hoping to not have to retrack it, but it looks like I'm going to have to do that anyway, so it seems like it might be best to resample to 44.1 kHz before even importing into Audacity and then do all the editing in Audacity while it's still in 24-bit format? Really appreciate the help on this.

If you did the tracking via labels, you can export the labels as a text file and re-import them later if you decide to make changes.

Without doing Audacity (combine, edits, export as 32bit due to edits & other processing) > r8brain (resample, dither to 16bit) > Audacity (track) then what you're suggesting makes the most sense. I've heard it's best practice to do edits in the largest domain possible (whatever your raw file comes in as, in this case 24/48), and resample, then dither at the end (in that order), but I'm not sure how much it really matters in the grand scheme of things versus resampling first.

If you can run r8brain as a plugin in Audacity (like as a VST or AU plugin), then you could do the resample/dither last, set your project workflow in the bottom left corner as what you set it to in r8brain, and (if r8brain will do it as a plugin) turn dither off. I've used Ozone from Izotope that way in conjunction with Audacity.

The bummer about all of this is it's a bell curve. When it's done perfectly, it's great (you don't notice any problems). When it's done sort of haphazardly, you probably won't notice any problems, or at least probably won't attribute them to resampling/dithering (because it's under the noise floor), but when you do it poorly, its sort of noticable during critical listening.

Yeah, it seems like you would lose anything you gained by resampling first if you resampled before any normalizing, etc. But I can't imagine there is that much different between 48 kHz and 44.1 kHz. I suppose if I had used 96 kHz. I think keeping it 24-bit though until the final export will retain most of the advantage? Sounds like optimally what you would want to do (especially with a 24-bit / 96 kHz recording) would be to pull it into Audacity and do all of your normalizing, compression (if you use that), etc. and then export it as 16/44.1 before splitting it into tracks? Then pull that back into Audacity and split / re-export with no dithering (if you can really turn it off)?

For this one, I'm resampling from 48 kHz to 44.1 kHz with r8brain first but leaving it 24-bit. Then I will do everything I normally do in Audacity (normalize, try toremove any offending noises, split tracks, etc.) and then export like I normally do. That should work?

I haven't been using labels to split into tracks (didn't actually realize you could). I normally put my cursor where I want the track split and select everything between cursor and end and then split. Sounds like labels may be easier. Appreciate all the help and feedback. If it doesn't work, I'll probably be back;)

[willndmb]

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

Be careful about dither. Audacity dithers regardless of change in bitdepth (or direction of bitdepth) if you've set it to dither on export. So unless you're flipping that setting deep in the preferences menu each time, you're dithering twice at 16bit.

It's a handy program, and I can't find anything better at it's price point, but it has a ton of quirks like that.

hmmm
i most likey am doing it twice then
not sure what kind of effect it would have doing it twice, but it sounds ok on playback
guess i will get in the habit of checking that out too

[page]

Yeah, it seems like you would lose anything you gained by resampling first if you resampled before any normalizing, etc. But I can't imagine there is that much different between 48 kHz and 44.1 kHz. I suppose if I had used 96 kHz. I think keeping it 24-bit though until the final export will retain most of the advantage?

Some argue that the difference between 44.1 and 48 is worth more to music recorders than 48 to 96, but the theory stands as to doing edits at the higher sample rate before resampling. Like I said, in the grand scheme of things, I wouldn't worry about it.

As for your bit depth, yes, you'll gain a lot from doing edits in higher bitrates until the final export when you dither to 16.

Sounds like optimally what you would want to do (especially with a 24-bit / 96 kHz recording) would be to pull it into Audacity and do all of your normalizing, compression (if you use that), etc. and then export it as 16/44.1 before splitting it into tracks? Then pull that back into Audacity and split / re-export with no dithering (if you can really turn it off)?

Basically yes (and yes, you can turn off dither, select "none" instead of a shaping or quality, but that applies until you change it, not per workflow/file).

For this one, I'm resampling from 48 kHz to 44.1 kHz with r8brain first but leaving it 24-bit. Then I will do everything I normally do in Audacity (normalize, try toremove any offending noises, split tracks, etc.) and then export like I normally do. That should work?

Sure. Technically it's the resampling that is your hiccup with the extra gaps at the end of tracks. So doing that outside of Audacity resolves that. (or hell, before export works too).

I haven't been using labels to split into tracks (didn't actually realize you could). I normally put my cursor where I want the track split and select everything between cursor and end and then split. Sounds like labels may be easier. Appreciate all the help and feedback. If it doesn't work, I'll probably be back;)

yeah, read up on labels. That's one of the things that really improved in 1.3.x land in my experience. You can even set them to cut on CD sector boundaries too. Like I said, it's a great program, just a lot of quirks.

the safest way i found is to do any edits like amplify in the original file
change the sample/bit rates
export
open the new file
track it out
export multi

Be careful about dither. Audacity dithers regardless of change in bitdepth (or direction of bitdepth) if you've set it to dither on export. So unless you're flipping that setting deep in the preferences menu each time, you're dithering twice at 16bit.

It's a handy program, and I can't find anything better at it's price point, but it has a ton of quirks like that.

hmmm
i most likey am doing it twice then
not sure what kind of effect it would have doing it twice, but it sounds ok on playback
guess i will get in the habit of checking that out too

You're just getting extra dither noise. When the band is raging, you won't notice, but if it's really quiet and you turn the stereo up you have a greater likelihood.

You're not the first who's done this, nor the last. I wish Audacity embraced dither settings like every other editor but they dont.