That’s not what I see when I do the spectrum on my end directly from audacity. Is there a weird setting in Audacity for the wav file output? Sample rate 44100, 32 bit float.
turned off Dolby on the laptop and this is what I get:
So that solves the upper frequency clipping.
It doesn’t solve the breath amplification after the 36Audiobook mastering.
I’m assuming applying a debreather in advance may fix that…maybe?
and not having to do them one by one would be nice.
Has revealed a fan noise in your recording space, (computer ? aircon ?).
Having a pro quality mic does not exclude those unwanted noises.
AI tool will de-noise, de-click, de-breath, de-plosive, de-ess, de-reverb, all in one go.
before-after AI processing
Whether that’s acceptable to ACX is another matter.
Yep. Laptop fan. Not likely to be avoidable. I can suspend it off the edge of the table and the sound sinks some into the closet linens, but doesn’t go away. It also comes and goes since it’s tied to how hard the processor is working–which means it’s likely to be there while I’m working. Since it’s largely a high frequency problem, would a low pass filter reduce it?
Which ai tool are you referring to or recommending? There are several.
36 Mastering has a Rumble/Earthquake/Thunder filter built in. 100Hz High Pass Filter if you’re counting. It was designed around the filter used in news-gathering and outside movie shoots.
ACX hates distractions. Anything that doesn’t sound like you telling a fascinating story over cups of tea.
They use a submission with absolutely clean pure zero background sound as the indication that you used effects, apps, or filters in your work. Dead air between words doesn’t happen naturally. Since chances are good you are not an audio engineer, they automatically reject the posting. That’s not arbitrary fluff. ACX produces many different products and services. It would be really messy having some of their products be damaged because the sound didn’t get along with an application you used.
ACX explicitly rejects the idea of using an AI Voice.
Gasping is still hard. I know of no good ways to suppress that.
Hire out. There are postings on the forum from people reading works written by other people.
Koz
There’s no way I would ever use an AI as the voice, and I think it’s important for me to read what I’ve written. It’s getting better, but I recognize it’s not perfect.
If there are additional AI tools or other processes that can address the fact that the 36Audiobook macro amplifies what was an allowable breath prior to the processing and makes it unallowable, then I’d like to try. I’m a stubborn old woman.
A few words about 36 Mastering.
The first thing the sound hits is the Rumble Filter. There’s nothing like having your production change volume in response to sound that’s so low pitch you can’t hear it. From personal experience, the first indication of an earthquake can be a lamp falling over. The event is making plenty of noises, but it’s so low pitch you, as a practicing human, can’t hear it. Your cat can, however. If your cat turns into a massive furball with big eyes for no apparent reason, an impending quake may be why
Next thing it hits is an electronic volume control. Really, that’s all it does. Turn the volume up and down so the RMS (Loudness) of the chapter comes out slightly higher than the ACX specification.
Then a peak limiter checks in to keep the tips and peaks of the blue waves below the -3dB value also required by ACX. Actually, being obsessive, I set the show to -3.5dB. Very slightly quieter
Remember the slight high error the volume control uses? The peak limiter almost always reduces the overall volume of the chapter very slightly. Couple that with the slightly high volume set by the volume control usually hits the ACX RMS (loudness) specification pretty much spot on.
At no time does anything change the noise. You can apply a very gentle noise reduction without triggering any production alarms. The shshsh or fffffff background noise between words is still there, but it’s quieter than the ACX specification. Actually (see: obsessive), I usually shoot for 6dB to 10dB quieter than needed.
This is a voice test I shot with that iPhone.
That’s what I sound like. Run ACX-Check on it.
Yes, I know there’s a gasp in there. I suspect if your air management is no worse than that, you may get away with it.
A fan that goes on and off also changes pitch slightly making it nearly impossible to suppress. I would probably place the laptop outside the studio and string good quality headphones back in. Headphones during recording is usually recommended anyway.
I had a picture of a stand-alone recorder here, somewhere…
Koz
This is a Zoom H1n stand-alone sound recorder fully outfitted for voice. Those are Sennheiser headphones with thick, heavy ear padding.
That’s one of my traditional furniture moving pads back there for echo and noise suppression.
I would kill to be able to say I got the paper towel roll at Piggly-Wiggly, but alas, I think I found them at Vons.
Koz
I appreciate all the help. You both have been very patient with me.
I think I see some of the problem. When I look at the overall volume in the raw clip, it’s not quite what it needs to be–both peak and rms–which means that the macro is going to amplify everything. That’s probably because I’m recording so early in the morning. I’m not using the full voice I would use in the middle of the day.
The unadjusted breathing isn’t that bad–not too different from what you’re showing in the last clip. The worst ones peak around -45 dB, which is about 10 dB softer than the breath you showed, but when it increases the volume on everything, the breathing goes with it and it becomes noticeable.
The fan I see in the multiplot shows up in that lower than 100 hz band so the 36audiobook macro is likely to cover it over, as long as the base volume on the recording itself is high enough in the first place.
The mic is positioned to my right and pointing at my shoulder, not directly at my face. The distance can be 6 to 12 inches depending on how far back I lean in the chair. It’s also pointing in the opposite direction from the laptop (where the text is). I suspect I just need to fiddle with the mic positioning and my own volume (not the mic volume, per se).
36 Mastering s not an on-the-fly compressor or processor. It does one simple volume change per chapter. You should not be bobbing and weaving during the chapter reading. That 6 to 12 inch spacing change can give you double or half the volume. Don’t move. This just kills people with two-hour long chapters.
Each file must be no longer than 120 minutes. Split longer sections into separate files and include a secondary header for continuity.
This is where headphones listening to yourself helps. Doubling or halving the volume is instantly obvious when it’s happening right in your ears.
I’ve always used a directional microphone, pushed to the side and aimed at my cheek. Lip smacks, clicks, pops, and breath noises go straight in front of my lips and as a rule don’t go into the recording.
I noticed you carefully avoiding using the word “headphones” in your posting.
Koz
Not avoiding it. Taking it for granted. Wouldn’t even try without headphones. At first I was using bluetooth headphones but when the bluetooth was connected up in the most fail-proof way, there was a latency between what I could hear and what I was producing that drove me bananas. I’m using a (cheap) over the ear wired headphone now with some pretty heavy foam that is plugged directly into the back of the microphone.
I was also assuming that if I changed the distance to the mic, I would need to rerecord everything to match, since acx wants it all consistent. The mic specs specify 6-12 inches, so I wasn’t being quite as picky as I should have. I’ve repositioned the mic so that it’s sitting right around 3 inches and dropped the pop screen (which was also driving me crazy), but not the mic foam. That should keep things a bit more consistent in terms of distance. The back of the chair is dead vertically straight and I’ve pulled the table up close so that slouching isn’t easy to do. I can hear nearly all movement, so I’m keeping that to a minimum and re-recording phrases if needed. Had to stop for a bit at one point because I could hear someone getting ice from the fridge.
The longest chapter recording is running 80 minutes or so before any text based editing and drops to 45-55 minutes after the edits. I’m doing my best to restart at the beginning of sentences or paragraphs with any errors. You’re right. Trying to fix anything at the word level is not ideal. Using a dog-clicker to mark the errors drives the ACX test algorithm crazy, so there’s no sense in even attempting to apply any modifications until the text edits are complete. As I’m cutting errors out, I’m doing my best to keep the spacing consistent with what I was actually speaking and leaving the breaths in. There are a few that are still louder than I would prefer, but they remain largely within the -45 dB range. The question of whether to adjust the louder breaths before or after your macro is still an open question.
Doing a quick check on the 36Audiobook macro within a clean section yields a pass on all counts (which I immediately undo, since I’ll be applying that macro at the end to the whole).
Feels like light at the end of the tunnel (which may not even be an oncoming train!)
That would be the recommendation, yes. The goal is not surgically correct sound, but something you can listen to without a headache and sense when your vocal volume changes. I did the headphone research thing but I suspect you’ve already got that sorted. My favorite headphones are the Sennheiser eh150. I suspect they were originally designed for disco listening because they have a bass bump, but I have no trouble listening to an entire movie in them as well as do production. I have two.
The “pro” headphones (Sony MDR7506) have an odd talent. They will show the sound lady on a movie set sound errors before the director perceives them. No, they will not survive listening to a movie. They sound hard and uninviting.
That’s what this is.
Now all we have to do is get rid of your computer. Put the script on your phone?
GASP Print it?
Koz
If I print it, then I’ll need to edit out the page turns–and I have plenty enough editing to do with just the unforced errors. Besides, my printer is on its last legs and the book is about 165 pages.
There’s an annoying tinny sound in my voice that wasn’t there when I was a younger woman, but I wouldn’t trade the years for the clarity.
I’d love a really beautiful set of headphones, but for the time being the silly ones I have will have to do. They really aren’t anything special. I use the wireless ones for editing–Bose Pearl, so respectable, but a flashier label than the headphones themselves.
Got done with today’s chapter editing, applied the macro and it passes acx with no trouble. I guess that means I have a working drawing board so I can really start recording. Third time’s the charm, right?
Truthfully, the goal was to get it ready for acx, but I know myself. Once I start reading it aloud, there were going to be changes, and the manuscript is basically ready for typesetting, so it will be probably a month or more before I can submit to acx anyway.
You all have been a huge help. Hopefully, the error edits don’t leave me with too much trouble from the checking team.
Do you have a heavy quilt/blanket/towel on the desk? That can help.
Last note, once you start reading, you can’t change anything. As noted above, all your chapters have to match.
Good luck.
Koz