Selection too long for analysis

Hi folks.

I’m pretty new to audiobook recording and just joined this forum so please be gentle!

First off, thank you from the bottom of my heart for what all of you, (whom I’m sure are paid handsomely for your efforts :wink: ) do to support this forum and community!! I certainly wouldn’t have had the relatively easy time of getting to know Audacity were it not for your help.

Recording on a laptop running Windows 10, i7 processor, 16GB memory, running Audacity 2.2.2, recording mono at 44100 bit rate. Mic is a SE2200 connected to a Focusrite Scarlett 2i2.

I downloaded the ACX Check plugin and so far after 4 books, have not had an issue with the files uploaded to ACX. It has certainly made processing my edited files easier.

My question is this…

Is there any easy way to modify the plugin for larger tracks? Many of the chapters I run it against run close to 90 minutes and when I run it, I get the “selection too long for analysis” error.

I’ve worked around it by running the check on overlapping sections of the track, about a 25 minute section at a time. While that works and is not really a painful process, I thought I would ask the experts if there is a better to perform that task.

Thanks in advance for any help you can offer.

Unfortunately there is no easy modification. The routines that ACXCheck uses to do the analysis require that the whole sample be in memory. Most of the analysis tools have this limitation.

Also I would say that if you have your mastering process down pat, there probably is no need to check every part of your work.


The ACX chapter limit is 120 minutes. I wondered.

Yes, there are limits to what we can do with the built-in programming.

Are you using Mastering 4 or have you designed your own mastering process? We get page after page of failures, so it’s good every so often to display a success story. Did you turn your broom closet into a studio? How did you manage heating/cooling?

Recording in a quiet room goes a very long way to making production easy

I’m particularly interested in how you use the 2i2. The 2i2 is a natural stereo device. Do you try to capture mono in one pass or do you capture stereo and clip off the single dead track later? If you are capturing mono, can you do a transmission sound test?


Quick note. If you are using Mastering 4, you don’t have to check everything. Just noise and you can do that with Analyze > Contrast, and you can do it on any length chapter.

Zoom into room tone. Drag-select a bunch of it > Analyze > Contrast > Foreground > Measure Selection. The reading sticks so you have to make sure to Measure Selection each time. We expect that to give the same noise reading as ACX-Check—or within a fraction of a dB.

If you have a different mastering process than Mastering 4, then you’re on your own.


It occurred to me long after I wrote that, that one of the tools in Mastering 4 may have the same show length problem. And that’s a shame because I think it’s the only RMS (loudness) management tool available.

I wonder if there’s an easy way to tell when the tools are going to poop out.


In many cases it’s platform dependent.
On 64-bit Linux, most Nyquist effects can manage very long files, subject to the amount of available memory installed.
On Windows, several Nyquist effects crash at around 2,147,483,648 bytes, which often works out at about 50 minutes of audio.

In many cases, plug-ins can be rewritten in a way that avoids holding the data in RAM. This is the recommended approach for effects that may be used on long files. New features continue to be added to Nyquist to make this approach possible for more effects, for example, it is now possible to get the RMS or peak level of a selection almost instantly without loading the audio into RAM.

For cases where loading into RAM is unavoidable, the plug-in developer would normally protect against the eventuality by checking the length of the selection before processing, and bailing if the selection is too long.

New features for Nyquist are mostly documented here: Missing features - Audacity Support though the documentation does tend to lag a bit behind the features.

OK. So now that we’ve been little or no help at all, I still want to do a sound test.

Set up for your normal recording process. Can you see the Audacity screen? Start a recording and slate it “volume and clipping test.”

Then say “wooooooooooof” several times and keep getting louder and louder (and closer to the microphone) until the 2i2 volume knob starts turning red. Glance over at Audacity and see where the blue waves are and how far up the bouncing sound meter goes. I expect the sound meter to be orange/red and the glowing red knob to happen at about the same voice volume. Slate it as you go. “WOOOOF! The knobs just now turned red, but the Audacity blue waves are only half-way up.”

You are doing a data overload test. Some stereo (two channel) interfaces do not like running single-channel mono. Some don’t care. It would be good to know which the 2i2 is.


flynwill -
Thanks again. Being pretty new to this audio narrating thing, I’m still paranoid that my files won’t pass the ACX robots, which is why I still check every file I upload.

Koz -
I’m using the legacy process, for the most part. I’m recording in my home office (spare bedroom). I’ve hung moving blankets on 2 of the walls, over the window and hang another over the door when I’m recording. The floor is carpeted and ceiling height is 9 feet.

I turn off the overhead fan and don’t record when neighbors are cutting grass, kids are playing right outside the window, the trash trucks are in the neighborhood, etc.

When I first started using the 2i2, I recorded in stereo then split off and deleted the dead track since I believed it gave me a larger waveform. Even with the gain on the 2i2 cranked up to where I was picking up a lot of background noise, the waveform still looked too small to me. I really didn’t see a lot of difference between the mono waveform and the stereo one.

Since then, I’ve moved my mic, adjusted my distance to the mic, turned down the gain on the 2i2 and just record directly in mono. When editing, I zoom in and change the track scale to zoom x 2 if needed, although I find that since moving things around, the waveform looks fine to me now with peaks aproaching 0.5 even before running any processing on it.

I’d be glad to “capture a transmission sound test” if you tell me how.

Regarding editing, unfortunately, I didn’t search the forum very well before I began recording so I kind of cobbled together some of the settings from here and my personal experience getting tracks to pass the ACX check.

I first sample my current room tone, then run noise reduction set at 4,6,3. I then sample an “outstanding” breath in the wave and run noise reduction again at the same settings. After editing, I then run a chain that contains the Equalization, RMS normalize, Soft Limit and ACX Check commands. If the ACX Check fails due to length of selection, I run it against overlapping sections and if the data from all the sections falls within 0.5 Db I feel pretty confident the entire track will pass.

I’ve been extremely lucky and so far have only had ACX reject a track due to length of room tone at beginning or end of track. I’m now finishing my 5th book and don’t want to change horses in mid-stream by moving to Mastering 4 until it’s finished.

Once complete, I’m planning to “start from scratch” and dial in all my settings based on your Mastering 4 process guide. Hopefully that will make the technical side of things even easier and I can focus on my technique and delivery. As a male on the far side of “mature”, I’m still having trouble doing female voices when required - go figure!

Thanks again to all for your help.

Koz -
As I was posting this I just saw your info about the sound test. I’ll try to get that done a little later today.

You should not change anything if you have a process that works. Voice, presentation and process tend to marry and if you change one thing, you may mess up something else. For one thing, Mastering 4 has no provision for mouth and other human noises.

I’m surprised that RMS Normalize works with longer performances. It should have the same problems that ACX Check does. They use the same type of programming.


The last couple of versions of RMS Normalize uses a new feature that allows it find the RMS level without loading the entire selection into RAM. The original version of RMS Normalize used the old method, and that has the same problem as ACX Check.

Hi Steve. Koz is prodding me to rewrite ACXCheck in this manner. I followed your link but didn’t see any obvious list of new features. Is there a list somewhere of Nyquist functions that are not memory bound?

Indeed I’m pretty sure such a check is built into ACXCheck. I don’t remember what the limit is but I remember that it was determined experimentally.

See the SELECTION property list:
Feel free to start a new topic on the Nyquist board if you wish to discus a new version of ACXCheck.