ACX keeps rejecting what Audacity shows is fine

[Voice-of-Amara]

I’m working on an audiobook, where the bottom track is a char who speaks in telepathy. I added the effects on the right to add an effect akin to mental speaking. Currently using acx check, the top one passes. The bottom track also passes, but together ACX shows that the peaks are at 0 and need to be adjusted.

I can’t seem to shift peaks with amplification without messing up RMS (too loud), using a limiter messes up both, and if peaks passes, rms fails and vice versa. I’ve gone through each number from -3–(-4.9) on peaks, and from -18-(-23) on RMS and neither are playing well.

It SOUNDS fine, no issues with peaking, silence, buzzing, or the like, but the automatic checking doesn’t like the complete file. I’m at a loss for what to do because this plagues all chapters with this character. I don’t wanna reverse the effect, as it’s integral to the story, but DAMN is this frustrating when I’m so close to publishing.

I even re-recorded each line in the bottom track, ran the macro on each line to make sure all of them passed acx check, and with both tracks passing (below) it still has 0db peaks when uploading it to ACXlab (document attached). I’m losing my mind and am in the final stages of editing and this plagues at least 12 chapters.

ACXLAB output for whole file

Track 1

Track 2

Macro Settings

I could use some help.

[DVDdoug]

I even re-recorded each line in the bottom track, ran the macro on each line to make sure all of them passed acx check,

What do you mean by “line”? You need to run the macro on the almost-complete file as the last step before compressing to MP3.

You shouldn’t be getting a difference in peak measurements. Are you checking the exact same files, after encoding to MP3, etc.?

Make sure to run ACX Check on the MP3.

MP3 compression changes the wave shape so some peaks can get somewhat higher but usually it’s less than 1dB of difference.

Sometimes you’ll get slightly different RMS calculations because you can’t practically calculate the RMS on on the whole file. It has to be chopped-up and nobody knows exactly how ACX does it

They will also reject you for dead silence. Either you added pure-digital silence or you use “excessive” noise reduction or noise gating.

That “overprocessing” should be avoided but you can generate some noise around -65 or -75dB (0.0005 amplitude) and mix it in. (Brownian noise is probably best because it’s harder to hear but should still measure “bad enough to pass”.)

ACX is probably measuring the average or RMS noise level so it may be lower than the (peak) amplitude you set in Generate.

[Voice-of-Amara]

@DVDdoug

What do you mean by “line”? You need to run the macro on the almost-complete file as the last step before compressing to MP3.

I mean that each little bit of audio on the 2nd track is a specific character’s lines. I’ll pull the extracted audio file back into audacity and run a check on that. Looking at it now, I see that when compressing to an mp3 file, it merged both tracks into one. Now as a single file, all audio effects included, it passes the acx check. Uploading it back to acx and…

It worked!!!

Alright. So then my production my process should include opening the exported mp3 back into audacity, running the mastery and checks on that file, and then having ACX check the file after I re-export it with the compressed and encoded mp3. This is only a problem on files that use more than one track, which I usually do when I need to add audio effects to a specific character’s voice.

I wasn’t checking after encoding to mp3, just opening up the aup3 file and being really confused. I will likely still re-record the lines in other chapters as some are over-processed, but this appears to have solved my issue. Thank you so much!!

[kozikowski]

A couple of notes. You should ignore MP3 until the final step where you submit. One of the ACX specifications for submission is that your MP3 must have a Quality of 192 Constant, or higher. If you make an MP3 from an MP3, the quality goes down in unpredictable ways—and you can’t stop it. I see ACX actually checks MP3 submission quality now. They didn’t used to.

Production should be in WAV or other perfect quality format and, as above, only create the MP3 as a last step. Also, your Archive Edit Master should be a perfect WAV or equivalent. Not MP3.

Have you submitted before? The Audiobook concept is someone telling you a fascinating story over cups of tea. Not a broad theatrical production or mixed podcast.

They also prefer mono over stereo. This is a submission sample I generated.

Yes, it does pass ACX-Check (attached) and ACXLab.

You can try submitting a multi-performer theatrical production with effects and music, but that used to be forbidden, even if you did pass all the technical standards.

This is a cousin to another problem. They hate distracting effects and distortion. I see you already ran into the problem of over-applying Noise Reduction.

I would try mixing the production down to two-channel stereo or, even better, one channel mono, apply 36Mastering, check it, convert to MP3 and submit.

Post back how it goes.

Koz

[kozikowski]

In going back over my notes, I think FLAC is the alternative to WAV format. WAV is uncompressed and FLAC is “Lossless.”

I’m checking that.

Also I’m reminded why they don’t much like music. Remember, it’s a business. Be prepared to present all the copyrights and ownership of your tunes.

Koz

[Voice-of-Amara]

I’ve gotten it to pass, and I’ve done two full books before. This is the first time adding effects so it’s an extra layer of mastering to get through. I’ll keep the WAV format in mind for the future, and I’ve done mono for previous works!

For effects, I’m only wanting an echo effect to differentiate mental speech from a character, from their eventual normal speech. previously I would simply speak quieter and closer to the microphone for the illusion of closeness. I got approval from the author for adding an effect that helps things be more distinct. I’m still finding I need to re-record the character’s lines, but I will export in FLAC or WAV, so that audacity merges tracks, then master it in a new aup3 file, and export as mp3. a pain, but a more thorough method that I may adopt moving forward should I do anything like this again!

[kozikowski]

@Voice-of-Amara

previously I would simply speak quieter and closer to the microphone for the illusion of closeness.

Works for me.

A note on 36Mastering. It doesn’t adjust volume as the performance progresses. It picks one value and applies it over the whole chapter. It was designed for one person reading a chapter with quality-control headphones and an even voice.

Koz

[ljkendall_author]

I was interested in this topic since I’m planning to create audio book versions of my own books, starting next year, and one of things I’ve wondered was how to make it clear when a character is thinking rather than speaking aloud. In the text, it’s clearly visible. For an audio book there’s no equivalent to italics.
I have no idea how you’d create a separate voice track for a telepath. It sounds like a lot of work, to record those little pieces of text separately!

[Wrecks0]

A trick I like to use on the podcast for “thoughts” is to record a sentence, reverse it, add a slight reverb, then reverse again. You need more silence than usual at the beginning and end of the selection because of the reverb. It creates a kind of, I don’t know, maybe foreshadowing feeling? Other people here probably have better ideas.

[ljkendall_author]

Sounds good. The problem would be finding and applying the effect to every thought in a chapter! I’ve been toying with the idea of some hand held object I could speak into or use to alter the sound as I’m doing it, to apply the ‘effect’ in real time as I’m narrating!

I suspect that may be too hard to get right. If the sound quality goes muddy or drops, readers wouldn’t like it I’m sure.

Maybe some almost inaudible musical tinkling sound in the background?

Dunno. I’d love to hear ideas though.

[Wrecks0]

For that, I would recommend looking for a K9 vocal processor.

[kozikowski]

One of the common faults of recording too close to a hard surface or an echo-ridden room is sounding like talking into a wine glass. You could get a secondary voice effect without massive effort or tons of software application by actually talking into a wine glass. Milk bottle works, too

The first author of this thread used to actually get close and quiet to the microphone to get this effect. Proximity does change your voice. My response was, “Works for me.” It probably works for ACX, too.

You can be wearing real-time headphones and get really good at auto level setting. You can do this without any post production.

There’s another note about effects. The objection that ACX has about having absolute dead quiet between words or sentences is not the damage that it might do to the production. It’s the Red Flag. People who submit like this are very likely to make other effects mistakes and damage. ACX hates distractions. You should sound like you’re telling someone a story over cups of tea.

Refer to the illustration I posted further up and read the Noise value. They want it quieter than -60dB. I use -65dB as a safety measure. Note the submission is -67.9dB.

Koz

[ljkendall_author]

That’s all super helpful information, Koz. I see the -67.91 dB in your illustration.

I googled 36Mastering: is this a correct summary of it?

“Audacity 3.6 mastering” refers to the new master effects channel in Audacity version 3.6, which allows applying effects to an entire project to shape the overall mix during the mastering stage. To master a track, export it, open it in a new project, and then use effects like Amplify (to normalize), a Highpass Filter (to remove rumble), and a Limiter (to increase loudness without clipping).

If so, I think I understand.

Many thanks!

[kozikowski]

Please note that a production will fail that dead silent noise check if you have “Generate Silence” of any duration anywhere in the show.

36Audiobook-Mastering-Macro.txt (585 Bytes)

There was an earlier version of that without the “36.” Audacity developers changed the application tools around in version 3.6 so ordinary Mastering stopped working. I’m not a fan of making serious changes and using the same filename. I think there are still manual notes that call the wrong macro.

One of the worst problems of this tool is that “macros” are ordinary text files. That causes all sorts of problems. Windows will try to hide the “dot t x t” at the end to “Help you.” So the application as used and the one in the instructions appear to be different. What could go wrong?

As in that explanation above, people feel it’s perfectly OK to rearrange the tools. It’s not. The tools talk to each other and they must appear as written and in the correct order.

The first step is the rumble filter. Low pitch thumps and rumble can cause your production to sound off loudness even if everything else works OK. The actual rumble damage can’t be heard, so it’s a little magic.

I didn’t write that. Steve the Fiddle did. It’s based on the rumble filters used in Broadcast and Hollywood field work.

Next is the loudness setting. Note that it’s wrong. It doesn’t set loudness in the middle of the allowed ACX loudness range like you would expect. It sets it slightly high so that on average, the next step, Peak Limiter will gently squish the sound down to the desired value. Note the goal isn’t -3.0, the actual peak limit. It’s -3.5 to allow for any conversion and file management damage.

I published a voice test from a cellphone on the desk in my quiet bedroom. No fancy-pants, expensive microphones and mixers. One pass through 36Mastering and it passes ACX-Check and it sounded exactly like me. Being obsessive, I applied gentle Noise Reduction of the Beast (6, 6, 6).

I’m on a new machine, so any minute now I’m going to find that sound file.

There is a joke about Mastering. You can put trash in there and it will come out passing ACX. So you do have to pay attention to your content. Also none of these processes have been checked with music.

Koz

[kozikowski]

This graphic and analysis appear a bit further up the message.

While it did pass ACX-Check exactly as I recorded it, the noise value was, in my mind, too close. So I applied gentle noise reduction. I note that the 6, 6, 6 value is now the Audacity Noise Reduction default value.

Koz

[3Guesses]

have you tried adding a new mono track and generating noise? I found YouTubes that have mentioned recording ambient room sounds and adding that, but when you add a track you don’t have to fuss over maintaining a file.
I tried to capture a screen shot, but I cannot seem to be able to post it. I opted for Pink noise at 0.001 amp and it puts the noise floor at just over 70. Then I save the noise along with the vocal track as one mono. Then reimport it and run it through the ACX check and you’ll have all three pass.

I hope that makes sense….

[kozikowski]

@3Guesses

I tried to capture a screen shot, but I cannot seem to be able to post it

The forum has marked you as too new to post content other than text. I don’t know what the limit is, but if you help out enough, the restriction will lift. We get a lot of people trying to advertise rug shampooing services in Bangalore on the forum.

@3Guesses

have you tried adding a new mono track and generating noise?

From the postings, I would say ACX doesn’t like the complex theatrical production posted (podcast) rather than a simple spoken audiobook. As I posted up the thread, I think getting close to the microphone and lowering your voice for the second actor has worked in the past and I suspect that will work this time.

For this trick, is highly recommended you monitor your live voice on headphones while you speak.

Koz

[kozikowski]

ACX is getting a little slippery about this, but my joke is the submission should be you telling me a fascinating story over cups of tea. Not cast of thousands, Hollywood special effects, and full orchestration.

Koz