Ableton Forum

I am cleaning up some HORRIBLY recorded audio for JAPANESE ESL tests in Japan right now. They were recorded in a studio with a shure SM57, and someone who knew NOTHING about audio. He blasted the mics, and had the recording gain set super low...so, I have TONS and TONS of noise. Noise can be cleaned up with Rx 2 advanced...but, since this studio was used by many people in between his sessions, the levels are different on each. I am using live to basically chop up mistakes that people made while readying 6-7 minutes of dialogue. I have about 180 of these...and am being paid for it...a decent amount, but NOT good enough to go into and really fix up everything. The biggest issue I am having is trying to get a decent normalizing level. Since all the recordings were on 1 SM57, and sometimes there are 3 people hunkered around it...and obviously different distances from the mic, one person is ALWAYS super louder than the others...throwing off the peaks really bad. And setting a compressor to handle this is a pain in the ass because everytime, I really gotta play with the threshold because the levels are different.

Does anyone have a good workflow to get some DECENTLY similar audio levels for 180 dialogues...kind of on the fly? I am an Izotope fan, and bought a lot of their stuff...ozone, rx 2, and nectar...and aint really up for buying any more plugins. Using abletons built in Normalizer wont do it because of the multiple speakers...and compressors are a bitch at the moment. Im thinking there has gotta be a STANDARD on bringing the levels to with a generic basic setting on a compressor that could handle this. As with the compressor, bringing up the noise floor...but after I will run it through RX 2, and take care of it. Any ideas!? This is my first time working with dialogues so excuse my naivety. Thanks guys!

FroBot

I don't know about a plug that would solve this mess instantly. But I had to do similar work in the past.

First do the noise reduction.
Then I would line up the audio in arrangement, split in chunks according to the level, then manually and by eye increase clip volume on the chunks with little volume so they all are at the same level more or less. Then run it through a gate and heavy compression.

i would run an auto batch of normalize process on all using wavlab or soundforge,

then apply a sligh noise reduce using a batch on all as well, then ear all results and make 3 folders relating process needz.

some might need compression due to the louder one speaking, some might only need noize redux ect, make overall categories.

then run again 3 or more batchs of deeper noise reduction/compression adapted for each category.

if you're not paid that much, go for batchs, don't enter in details, that would cost too much of your time.

if no wavlab or soundforge, you could sort samples on 3 or more live tracks, test the fx for each tracks,

then use freeze/flatten as audio to automatize each sample processing.

Would wavosaur or audacity allow batch processing this way?

TLDR. do we need to know all that?

'mp3gain' and 'Switch' are two tools I've used in the past that do batch audio processing.

UncleAge wrote:Would wavosaur or audacity allow batch processing this way?

http://manual.audacityteam.org/index.ph ... Processing

Yea, I have though about these routes...but I really dont think I made my point strong enough about HOW BAD the peak are different. We are talking 30 db in difference somtimes...it really is HORRIBLE...and using RMS mode...well, the middle is going to be drastic for both.

We have similar kind of situation at our work. Thousands of artist interviews from the 60's to today.
Some recorded with shitty tape recorders, some with shitty MD recorders, some with decent ZOOM H-4's etc.. Some have 60min caps when people left for coffee.. of course with the rec on. Everything digitalized with crappiest of gear (creative sound card -> audacity -> done) by people who know absolutely nothing about audio.

I said NO WAY! when they suggested me for cleaning those up

Sound Forge. I used to do a ton of voiceover recording, it's easy to set up a job that will load everything up, trim the silence from the ends, apply a de-ess filter, and normalize.

Mac user here though.

rosti=smart guy

I'm with loydb, I keep a windows box for stuff just like that. Which is infrequent.

It may be worth picking up a copy of Win and Soundforge if you have an intel mac, just bootcamp it.