I received a voice recording (here is a sample from a larger file)
I don't understand what caused the issue, and I don't understand how to make it tolerable. Trying the de-reverb treatment in iZotope RX10 yielded no results. Is there a solution?
-
It's all a bit muffled & bizarrely one side of a stereo file [which doesn't help] but once spilt to dual mono De-Reverb does a half decent job of it. You could probably do with some spectral shaping to rescue the dull thumpiness a bit more.– TetsujinFeb 2 at 12:53
-
Yeah I got like 5 files which are all left channel only recordings, perks of being an intern. What settings did you use on de-reverb? And I don't exactly know what you mean by spectral shaping. Could you clarify?– salmonlawyerFeb 2 at 13:01
-
I just told it to learn, then lifted the overall amount slider. I didn't save anything, just a 2 min look at it. Spectral shaping - like EQ, but you have to 'invent' higher frequencies, because the recording is as dull as all heck. Try to persuade your recordists to sound check first, or take at least an elementary sound recording course ;) RX has spectral repair & recovery plugs.– TetsujinFeb 2 at 13:04
-
So what module are you using to do spectral shaping (unless you mean to use EQ and just raise up the highs)? And these are recordings made at home so there is no technician to blame. I literally got these as MP3 files.– salmonlawyerFeb 2 at 13:08
-
Either de-reverb or dialog de-reverb will do it, first needs to learn, second will just work at defaults. I guessed these were amateur recordings [a sound engineer wouldn't have a job for long with this output quality] hence my recommendation for your 'recordists'. They need to learn how to record… & how to record in mono.– TetsujinFeb 2 at 13:10
|
Show 1 more comment
1 Answer
It's peculiar - the tops of the vocal bits are chopped off in the spectrogram in very deliberate chunks. I wonder if that's a file compression thing.
But anyway, first, I'd remove the noise (learn from the biggest "silence" gap), so any subsequent processing doesn't also process the noise. I'd try to be heavy on the gating, as it seems no heavy spectral cleansing is needed (it's also different during the vocal windows, like that compression you get on phone videos is present). Although not essential, you could also attempt to smooth the sharp 16 kHz roll-off simultaneously with the harmonic synthesis slider.
Then I'd de-reverb. There should be an early reflections preset (it won't only process early reflections). Watch out for noise removal artefacts though - it just needs to be subtle. If you don't like too many settings, use the dialogue de-reverb with the light reverb attenuation preset.
Then, if it's all good, I'd start on the enhancements, like EQ, etc.
Patented n00dles Tip: When you're using the de-reverb/dialogue de-reverb, invert the amount slider to get a feel for what you're actually targetting (It's similar to using the "output noise/reverb only" switch, but different). This inversion technique is useful in a lot of trial-and-error processing situations.
-
1There are a whole slew of highly-resonant bands in the mids & lows - presumably caused by the room - & the audio does go right up to 16k which should be plenty for spoken word… but the consonants, though suppressed, are actually badly sibilant as soon as you push them [you end up having to de-ess it as well]. I've tried a few methods of 'fixing' it [none came out well], but I pity the guy who has to try get this broadcast-ready. If they've got the intern on a task like this, when a pro would throw it right back to the client to re-record, it's a sorry state of affairs.– TetsujinFeb 2 at 18:41
-
1Thank you for the advice, as I said in the comments on the question this was sent to me as MP3 and likely recorded on a phone app without regards to any settings, so that's most likely why it falls off sharply at 16KHz. @Tetsujin this is going on an app so thankfully not a broadcast situation but yeah I told them the moment I got these files that we'd need to re-record; as usual "try to fix what you can" is the mantra. Feb 3 at 5:04
-
1@Tetsujin (I deleted my comment by accident, xD I'll try again -) Yeah, you're probably right (I didn't push it at all). I think they just want it to be presentable to the layman for now. I've heard worse in apps, ads and even on TV on occasion. For the HF, I don't like unnaturally sharp roll-offs, so just wanted to smooth it a bit, but the harmonic synthesis can go bad quick.– n00dles ♦Feb 3 at 12:36
-
1salmonlawyer - I think just sometimes you have to push back on these kind of instances. Do it once & they'll think you can fix anything, any time. ;))– TetsujinFeb 3 at 12:41
-
1@Tetsujin yeah I used Adobe Podcast and was quite impressed. I didn't even have to Mono the audio, I just threw the file as-is into the service and it spat out a quite good file. There are still artifacts there, but wow. I'll have to keep this a secret from the bosses though, don't want to get the artists into the bad habit of doing poor recordings... Feb 4 at 16:11