Working on my first audiobook. Just got a 2nd notice that I’ve failed the noise floor.
I use the ACX plugin and my files have all passed individually when I check them, yet I still fail when ACX runs their check. I’m a new user here so I can’t upload any of my audio files here, so that kind of cuts me off at the knees. Can anyone suggest what more I might do, or look for, to resolve this?
If you include any true -∞ (minus infinity) silence in your upload to ACX that will automatically trigger the “noise floor too low” message.
Two ways -∞ can happen: #1: Your computer/audio-interface has applied a noise-gate before the audio reaches Audacity. #2: You have created/added true -∞ silence when editing the audio.
There’s a free audio-host associated with Audacity: audio dot com …
Thank you so much for answering. I realize this topic is brought up a lot.
Ummm, I’m starting to suspect that using the spectograph to delete random noises to make the areas between sentences quiet (blank on the spectrograph) was a bad idea. I wanted perfection. Dear lord, do I have to re-read the entire novel? It would serve me right, I guess.
Thanks for the audio.com tip! I’m including some samples now.
i.e. it’s the same quality as voicemail/Skype.
Audiobooks have to include sound up to at least 16000Hz, i.e. CD quality.
Heresy warning:
I would not use the latest version of Audacity to make an audio book.
I would use its free competitor OCENaudio, (or an old version of Audacity).
this was immensely helpful! thank you. I used OCENaudio, but had to use audacity for the ACX check. this sample passed and I saw the sound was near 17000Hz, so that makes me hopeful. I haven’t edited out breaths or anything else in this sample, I’ve just tried to get it to ACX standards to make sure I’m doing it correctly.
That has the full audio spectrum . However …
it has conspicuous noise-reduction artifacts, (and a noise-gate).
The presence of either of those causes rejection by ACX,
The noise-reduction & noise-gate could be being applied before the sound reaches Audacity/OCENaudio, by “audio enhancements” which should be turned off for a faithful recording.
I don’t know the “audio enhancements” can be on a Mac,
( I don’t speak Mac). I can only point you to the Audacity manual …
"Mac
Navigate Apple Menu > System Preferences > Hardware > Sound
Choose the “Input” tab and ensure that “Use ambient noise reduction” is unchecked. You may have to select the built-in internal microphone to make this checkbox visible. If checked, it may apply effects to external microphones as well as the internal microphone. "
I did the noise reduction, but I don’t know what the noise gate is yet and I read that ACX doesn’t allow files that have used it. So noise reduction isn’t ok either?
Also, what would be the best method? Record on OCENaudio and do all of the sound stuff using audacity?
If it’s obvious that noise-reduction has been applied, then it fails ACX …
If Audacity was as-advertised, then it would be the better of the two.
However recently Audacity has become sluggish, unreliable, and creates bloated project-files. I doubt an Audacity-3 project which contained many hours of audio (an audiobook) would be usable, or avoid corruption.
[ Corrupted project => all your work is lost ].
Whereas OCENaudio is reliable, fast, and uses a fraction of the amount of memory & CPU used by Audcaity-3. Unlike Audacity, OCENaudio comes with a real-time equalizer, compressor, limiter, (i.e.no plugins required).
However recently Audacity has become sluggish, unreliable, and creates bloated project-files. I doubt an Audacity-3 project which contained many hours of audio (an audiobook) would be usable, or avoid corruption.
The sluggishness is likely mostly related to waveform rendering performance being heavily dependent on window size. If you use an older monitor, it’ll feel much more responsive, so that may be why Audacity 2 felt better to you.
Bloated project sizes are caused by applying effects destructively, and by non-destructive trimming. The former was also a thing in Audacity 2, but hidden better as the piles-of-files obscure it, the non-destructive trimming can be worked around by copying the project to a new window and only importing the visible portions of the clips. Rendering the track also gets rid of them.
As far as I can tell, corruption generally doesn’t happen on its own but are mostly caused by cloud syncs, or other processes which interfere with the creation or deletion of WAL files. The size of the project doesn’t really matter here; just yesterday I saw a user with a 5000 hour (sic!) project. SQLite is fairly resilient against data loss too, and most often corrupted projects can be recovered afaict.
Additionally, you can use audacity “ocen-like” by exporting as wav every time rather than saving projects.
I’m not aware of major reliability issues, please report them in the bugtracker if you have any evidence for them. Crashes and data loss get high-priority fixes.
can you address the audio from the link I sent? using audacity gets me maybe 12000Hz, whereas OCENaudio got me to 19000Hz. I watched videos of how to make it live up to ACX standards, but now i’m being told that ACX won’t accept if i’ve used the noise reduction feature. how do people at home do this?