I’ve tried noise removal, diagnostic for click and clip, high pass of 175, deesser, desherer, breathing plugin, plosive plugin (thought it could be added to the entire track but had to be added to each plosive) and still left with lots and lots of plosives especially on the t-words as twice, two, tactic, time…
Do I really have to go through 40 minuttes of plosive removal manually, lowering amplify on the plosive wave, or is there some miracle plugin or fix that will do the work for me?
I’m using Adobe Audition but have access to Reaper and Audacity if there are some miracle plug in for those apps.
This is a snippet of the original (mp3) file:
Hope you will be able to guide me in the right direction.
The golden rule for audio is to get it right at source. Sometimes this isn’t possible but you have to always aim to record as close to perfect audio at source.
Post processing should be to add a polish to the audio. Anything other than that will have an effect on the overall quality of the audio. For example running noise reduction or denoise will reduce the overall quality of the audio. I didn’t programme audition but it stands to reason how noise reduction works is by capturing a noise print and removing the frequencies within the noise print which may also cross with frequencies within the audio you’d like to keep but will be removed.
Traditional plosives usually occur within the lower frequencies and can be removed with a bass roll off. I’m just listening to that on my iPhone but it doesn’t appear to have much low end frequency anyway. We can sacrifice at least up to 80hz and unless you have an unusually low and boomy voice the human voice doesn’t contain frequencies that low.
From what I can hear on my phone, and it may well be different if it was listening on my monitor speakers the audio is destroyed here by higher end problems. Plosives usually come with hard D, P and B sounds. Your b sounds are clear in that audio so this isn’t ruined by plosives but by sibilance. A traditional plosive can easily be detected by holding your hand in front of your mouth and saying Dog, Ball or Plosive. You’ll feel the breath hit your hand as you say the words. You can get a pop filter which helps reduce plosives, some microphones are also less susceptible to plosives but you need to work on mic technique to reduce the breath and to slightly turn from the mic possibly so you’re breath isn’t directly hitting the mic.
Audition has a native de-esser which may be your only hope of saving this audio. What I would suggest is to open the de-esser and play through the audio. Click the output sibilance only box and sweep up and down the centre frequency until you hear audition output the noise you want to remove. Then uncheck the output sibilance only. From there you may also have to adjust the bandwidth to cover all the sibilant area in that audio. You’ll also have to tweak the threshold as well until you get a result you’re happy with.
I don’t think you’re going to make this audio sound anywhere near perfect but I think you can make it sound better.
I’m not sure what mic you used, or in what environment you recorded this but the less post processing you have to do in future the better quality audio you’ll get. Thing of post and a final stage, normalising to a good level, adding a little polish in terms of EQ. For what I believe you’ve recorded here you wouldn’t want to be doing too much more otherwise it will sound over processed for what I think you’re looking to achieve. Obviously if you’re making imaging then post-processing goes further by adding compression and vocal effects but that’s a whole different sound to what I think you’re looking to achieve.
I think going forward you need to look at the mic you’re using. Some are more sibilant than others and with the sibilance in your voice you need to aim for a microphone less sensitive to the sibilant frequencies in your voice.
Thanks for replying.
I didn’t record it, it was handed to me this way. It’s a skype interview without proper mic.
The clip you are hearing is the original, not the edited piece, I did run it through noise processor and I did add auditions native deesser and desheerer, but if what you’re saying about the t-words being a problem in the sibilance, then a tweaking in the bandwidth and threshold in the sibilance might just be the cure. Thank you so much, will try this out and see will it help.
The thing with sibilance is that everyone’s voice is very different and sibilance (if it occurs) occurs at different frequencies for everyone. It might be possible that someone is sibiliant at a frequency a microphone really picks up.
I actually know a lady who is audibly sibilant without even using a microphone. Most people you wouldn’t notice the sibilance until you record them but this lady has (to me as someone who’s spent years working with audio) got a voice that’s sibilant to the human ear. I can’t imagine what she’d be like on a mic!
Good luck with this
Didn’t really make much difference, still loads of disturbance in firsttttt tttttime intttttroduced dojjjjjo ttttwo. Just can’t get it to sound right without castrating his voice entirely and that doesn’t even remove all of it.
To be honest I think that audio is beyond repair. If you were to go through the spectral frequency with the spot healing brush you could have a stab at improving it but there’s so much to remove the audio would be awful after.
I think the only thing to do here sadly is get it re-recorded to a better standard and explain to the original recorder that there’s only so much post production can do and that audio is beyond salvageable.
I went around this the long and hard road.
Every time a scchhh or tttt was hard on the ears, I lowered the amplitude for that particular sound in the spectral view. I found that the spot and auto healing tool are too drastic and leaves a distorted after “taste” hence the long hours, matching the harsh sounds with the digestible sounds.
It’s not perfect, but it’s easier on the ears and will do for an online podcast audience.
Thanks for all your suggestions.
Just out of interest, I dropped the file in question into an RX7 demo version, pressed ‘Repair Assistant’, and then selected one of the options that RX came back after a quick analysis… Not perfect, but I could work with this…
See what you think:-