How to save audio ruined by fan at memorial service?

[ehbowen]

I just came back from a memorial service for the mother of a friend this weekend. I had my (Nikon P610 consumer point & shoot) camera and tripod with me, so I asked her if she wanted me to set it up and video the service. She said that they already had a camera set up, but that I was welcome to do so.

Long story short: Due to poor placement and other factors the video from her camera was unusable, so she asked if there was any way to use mine. However, the service was held in an un-airconditioned barn and the owners had set up a noisy three-bladed fan on high speed not far away. Plus, they were not using professional-quality audio equipment and no audio recording was made other than what I got on my camera.

I've tried using the high-pass and noise reduction features of Audacity on the audio track and I've had some notable improvement, but I keep losing desirable sound. This is my first real effort working with Audacity's effects and I'm sure I've got a lot to learn. I'm attaching three clips: The first is of my mother playing "Blue Moon" on an electronic piano, the second is of my friend singing a tribute for her mother, and the third is of the pastor (her husband) and his opening remarks.

No, I don't expect miracles, but I'd like to do the best job I can considering the limitations I have. I can obviously upload longer stretches of the audio including room-noise-only clips if it will help. The video is HD and it obviously takes up a massive amount of memory; it really needs no help. The audio here is the problem.

[lordsmurf]

I don't want any video.
I need the entire audio file as WAV.
Upload to Dropbox, Google Drive, MS OneDrive, or iCloud.
PM or email me the link.

I can't do audio work during the day due to AC until winter. So audio at night, when cool, no AC. I'll try to do it tomorrow morning. I did a quick test, until the AC kicked back on. This isn't a lost cause, and some good improvement may be possible in Sound Forge.

[ehbowen]

PM sent. Thanks so much for the offer of help; it's more than I have any right to ask for.

Edit To Add: Please note that the audio which I uploaded, both in this thread and via link in the PM, is the original raw audio as recorded by the camera and does not include any of my attempts to massage it in Audacity.

[lordsmurf]

It's really easy to remove all noise, but then you get tone loss and artifacts. The worse the source, the easier to have those problems.

See what you think of the PM'd file.

I quickly ran an auto inverse NR, then manually pulled it back, took 4-5 attempts.
Then a Restoration, slight tweak to one of my existing filters/presets.
Then the graphic EQ, ran my Mono VHS4.
Tried some more, but nothing really helped, or did more damage, so just stopped for now.

BTW, I have a pro camera, and it also has 20-30 minutes file limits. One of the frontiers of still camera video has been to push those limits AND increase both fps and resolution while doing it.

[ehbowen]

That sounds pretty good! Certainly better than I expected considering the source limitations. For the other two files, do you think that I could come close to duplicating your efforts with the $99 SoundForge Cleanup package, or were you using personal filters/settings that this wouldn't give me access to?

[lordsmurf]

This is 100% using my own custom filters/presets, and I don't know if the Audio Cleaning Lab would help (doubt it, but downloading a trial for myself).

I could probably do a much better job on vocals, but you need to understand something: you cannot run audio filters "per video" for something this variable. It has to be treated like clips. You have music and singing, and the talking. Not just talking, but talking by different people that all seem to have different loudness levels (due to different microphone levels, ie standing too far away or too close, too short or too tall). Each segment/clip must be edited into a video file, and WAV made from each video file. Only then can it be restored well.

For example, on video1, you have music+singing, a guy, a women, another guy. I can make the music sound good with one filter set, and 1st guy sounds pretty good with another set. But running those same filters on the woman is horrible. Her voice and speech has a quality to it that blows out vowel sounds, and the "guy filter" run on her makes it blow both your speakers and ears. I'd bet that 2nd guy is tall and too far from the mic. Furthermore, his deep-south accent (lack of enunciation, dull speech pattern) will give any filters trouble, period.

Note that I do not want to randomly guess at breaking up the long clip. You really need to take visual cues into account for the breakup. Plus it will help later when editing, to prevent sync issues.

So ... clip it up for me. Video1 should have at least 4 segments.

Audio doesn't take me too long because I've also been doing it for more than 25 years now. That, and the fact that even the most complex of filters just takes a few minutes of processing time these days. Getting it to enjoyable quality isn't usually hard for me to do, and it's just that last mile to perfection where skill get harder (sometimes I can, sometimes I can't). I can tell this is important to you, so I don't mind a few minutes here, and a few there, too see about decent rough improvements.

[timtape]

My first post here. I've worked in audio for many years. Fan noise like that is perhaps the most difficult to deal with because it's composed of all audio frequencies at once. It's the audio equivalent of a small snowstorm between the camera and the scene we're trying to shoot.