Looking for feedback on my application rejected for over-processing
Hi community - my second application was rejected for over processing and I'm hoping that the Bunny community can help me zero in on the problem. I'm using a Lewitt 440 Pure mic, and applied only the following: 1) light de-ess & light mouth de-click in RX-8 (my first application was rejected for harsh sibilance and mouth sounds) 2) high pass rolloff at 75hz 3) manual editing of breaths in spaces between words (hand-reduced gain).
https://soundcloud.com/rersek/part001-revised
Much appreciated!
Please sign in to leave a comment.
Hi Richard, I had a quick listen and if you concentrate on the part where you say 'counting on you' in particular, you can hear a lot of processing in the form of high frequency noise. I'd advise having a listen and one-by-one, removing each type of processing until you find the one that's causing the problem. Happy to listen to the next version if you'd like me to and I hope that helps.
Thanks Dave! I’m away from my studio for a few days and will check this out and post an update next week
Hi Dave,
I think I could hear the high frequency artifacts you were referring to and made adjustments. As far as I can tell, it's better. A bit difficult for me - I have hearing loss and had to crank up my hearing aids a bit to zero in - so I'm hoping I caught it.
If I got rid of it, is there anything else in there that might not pass QC?
Here's the new file: https://soundcloud.com/rersek/part001-revision2
Thanks,
-Rich
Hi Rich, I had a listen and it does sound a lot better. However, I can still hear some distortion around the same phrase...'counting on you'. The best way I can describe it is 'high frequency phasing'. It's the sort of effect you get if a file has been compressed heavily so I would think that Bunny might view it as over-processed still.
Which elements of your processing did you remove/adjust this time around? It would interesting to hear it without the RX-8 processing as the high pass and manual de-breath are very standard so shouldn't be an issue.
Hi Dave - I could hear and see in RX8 some of what I think you're describing in that phrase between 3.5k and 5k if that's what you're referring to. I did some backtracking and it's in the original unedited recording also. Here's a link to the unprocessed file (includes edits but nothing processed besides that): https://soundcloud.com/rersek/part001-raw
Based on this, I'm going to work on the recording environment. The Lewitt mic comes with a metal pop filter that is held onto the suspension cage by magnets. If that were to resonate a little in the connection, it might be around that frequency. I'll try my old pop filter instead to see if that makes a difference.
Back to your question, I made some changes in how I was de-essing. I tried using the 'darken sibilance' setting for example. You can hear the sibilance in the raw file I just attached - I don't think that would pass, especially since I'd already gotten flagged for it in my first attempt.
Morning Rich, yes I can hear it in the raw file as you rightly say. I too use the Lewitt 440 but I didn't think the metal pop filter was very effective as it was too close to the diaphragm. I've used a simple cloth one instead from day one.
Out of interest, how is the rest of your recording chain set up? What's your recording process?
Thank you for confirming that, Dave! The recording chain goes from the mic to a Focusrite Scarlett Solo 3rd Gen (AIR mode on), to a Microsoft Surface. I use the ASIO driver, going into Tracktion Waveform DAW. I'll make edits, then run through my rack (high pass, de-ess, de-click), and render the final.
That all sounds like it should produce great quality. There is a possibility of course, that it's just the way your voice works and maybe a different mic technique could help.
If you aren't doing it already, don't speak directly into the mic. Aim the mic at your mouth rather than aiming your mouth at the mic. Speak 'past' the mic and see if that helps with the air and 's' sounds a bit.
About 9" from the mic.
I don't want to 'teach my grandmother to suck eggs' as we say because you're probably already doing this the right way, but just in case I thought I'd better mention it.
There could certainly have been something rattling around in my vocal chords - that happens sometimes and a little warm tea helps. I record off-axis from the mic but will see if I might be able to improve anything by addressing it a little more off-axis, etc. Thanks for your help! I'll take another try at the bunny soon!
-Rich
Thank you - these are great recommendations, Hiroshi! I've already been trying a new off-axis position for my mic. Turning off AIR mode makes good sense also - that should darken it up as well. I didn't think about using the windscreen though until you suggested it. I usually have avoided it due to the tendency to cut frequencies, but... that's maybe just what I might need to cut sibilance.
I took an entirely new shot at the recording with the recommendations - here's the revised recording: https://soundcloud.com/rersek/part-001-092321
Thanks,
-Rich
Hi Rich, I've had a listen and I can hear quite a bit of background noise. High frequency sounds like rustling of some sort.
Could it be the clothes you're wearing? Might be worth just doing an experiment recording that has a few seconds of silence somewhere so I can hear your background noise clearly.
Hi Hiroshi and Dave,
Thank you for the excellent feedback! I went on a quest this weekend to re-listen and find the issues that you've identified. I found it useful to select spaces between words and phrases and increase the gain about 45db. Makes it much easier to hear the flaws. I started identifying possible sources of the humming that you both heard. I record in a decent size walk-in closet with lots of clothes and thick carpet & padding but the possible sources could have been: heater blowing air through heating ducts in the room outside the closet, high- and low-frequency noise from either a LED light fixture in the closet, the power cube for the iPad I use to record, or the power conditioner. One by one, I recorded each in isolation and found that the culprit was the heating system (gotta remember to turn that off when I record). Then there were the rustling sounds - that could have been me moving my feet on the carpet, moving my arms around, or possibly the copy that I was holding. I attached the copy to a clipboard this time, and made a conscious effort not to move a lot.
I was using the RX de-esser on the previous sample but tweaked it to reduce sibilance further (hopefully without sounding over-processed). I'm now using the slow response rather than fast response setting - that seems to maintain clarity a little better.
Here are the results - when I use the '+45db' test, I'm not hearing the junk I used to. I'm posting two versions. The first is with nothing more than a little de-ess, and a 70hz high pass, plus some hand-editing of a few mouth clicks. The second is an experiment with the RX voice-denoise noise gate (adjusted to just barely cut noise) - is that taking too much out of the audio? Which sounds better?
Thanks for all of the help!
-Rich
https://soundcloud.com/rersek/9-24-21-part001
https://soundcloud.com/rersek/9-24-21-part001-with-noise-gate
Hi Rich, these are so much better than your original ones. Well done for persevering. As far as comparing the two, I slightly prefer the first version, but only slightly.
It's amazing the difference clothes can make. I have 'noisy' t-shirts and 'quiet' ones and I didn't think that was a thing to be honest.
Best of luck!
Thank you so much, Dave and Hiroshi! Looks like it's time to take another crack at the test :-)
-Rich
Hi Hiroshi,
I took a look at your sample in RX and it looks like the editing fades out the spaces between phrases and fades back in. Anything more? I was hesitant to do that because I thought that the QC team might flag that kind of editing. My goal was to leave in the room tone (so long as the noise floor is low enough). Should I be rethinking that? I do think I could edit those spaces with a low pass / high pass filter with something like 12db/octave rolloff. But would be interested in what the community generally does there.
Thanks,
-Rich
Hi Rich, I use Reaper for my VO. (admittedly I do very little actual work but I did learn how to use it properly for when I do need it).
I have my VO track and I have a real-time process going on which has three parts:
I also have a track that is nothing but room tone (can be extended to match the project length), which measures -60dB.
This track also has a 100Hz high pass and a 18KHz high shelf.
The trick here is that this track has an inverted gate that opens when the VO track hits a particular low level and allows the room tone through. It's basically 'Ducking'.
It's all automatic and works in real-time and the attack and release rate are all set as a preset to sounds natural, so all I need to do is manually drop breaths and mouth clicks.
Then I create a stem that has the fx all applied and is normalised at -3dB (again, an automated process) and export that as a WAV.
Hiroshi,
Thank you for those insights - it's definitely something that I was wondering about and so as you and Dave suggest, that would help reduce some background hiss.
Dave,
Thank you for sharing how you manage your workflow. For something small like the sample I provided, it was easy enough to do those kind of edits by hand in RX but I might try experimenting in my DAW with the sidechain replacement technique you're using. I think I'd make sure that's working well before I submitted anything that way but I can see how it would work and save time.
Thanks,
Rich
Thank you, Hiroshi! I'll make sure to select that.
-Rich
I came back to say 'Thanks Again!', Dave and Hiroshi! I took the application test this week and passed!!
That's great to hear Rich! Well done for persevering :)