Sound files: advice and examples

I uploaded my first audio observation to iNat here. I used some tutorials on improving audio quality of bird sound recording for ID purposes. To my untrained ear, it sounds ok and recognisable, but I’d love to get some feedback from people who are more experiencing in recording birds or listening to audio of them. Does it sound good enough?

What are some good audio observations on iNat? Which ones have you found particularly easy to work with?

Thank you!

To my ear, it sounds overly processed, such that there is no background noise at all, and perhaps as a result, the sound seems a little distorted. For ID purposes, it should be fine, but I’d suggest to crop and amplify as necessary and not do any other processing. I found this tutorial very useful: https://www.macaulaylibrary.org/resources/audio-editing-tutorials/editing-in-audacity/

4 Likes

This is not an owl or region that I familiar with, but I think that is good to ID the species. It sounds quite filtered, though, muffled, missing upper registers. With many sounds, I find my ability to readily recognize the species is dependent on hearing overtones.

Other that trimming, especially to get rid of jarring loud incidental noises in otherwise quiet recordings, unprocessed or only slightly amplified audio is usually my personal preference. A listener can process for themselves, but once filtered out, those subtleties cannot be returned. I believe having both the processed and unprocessed audio is also an option within guidelines.

Some notes are distinct by themselves and even a one second recording is good, but longer recording with multiple types of call notes, multiple song renditions, or demonstrating their spacing can be very helpful with others.

3 Likes

I have a well-trained ear but not necessarily with bird calls, so this is just general input. I’d recommend editing a lot less if you choose to edit at all. Trimming the audio to where it’s most clear might work better if there’s a lot of background noise. This audio kind of sounds like somebody put a bird in a jar, because it’s so contained. Most of the time (hopefully all of the time) bird calls aren’t coming from inside a jar, so leaving it unedited might sound less appealing if it’s a little messy with background noise, but it’ll probably be more accurate to what other people typically hear themselves when listening to birds.

4 Likes

it sounds like you’ve applied too much noise reduction to the recording.

if you want to clean up the noise a bit, the way i would do it for bird stuff is:

  1. open up your sound file
  2. select a section that’s just background noise
  3. analyze the frequencies (plot the spectrum) for that section, and look for peaks
  4. select your whole file
  5. for each major frequency peak above about 2600Hz, apply a low pass filter, which will get rid of extra hiss
  6. for each major frequency peak below about 2600Hz, apply a high pass filter, which will get rid of low drones and hums
  7. if needed, increase the volume (gain). if i do this, i would usually target audio peaks between about -20 and -12 dB or thereabouts, or i might go softer, if the audio is really ugly.
  8. trim, if needed.
2 Likes

Thank you! I’ll definitely update it with a less edited file. When I changed the volume (as per tutorials) there was so much background noise that even running the tool to remove it just once or twice didn’t seem enough, but I know I know.

@pisum, unfortunately, I don’t understand what a lot of that means. I used the normalise tool on Audacity set to -3 db and then filtered background noise, but I’m not at all a sound expert. Do you have a tutorial for something like this, by any chance?

This was one of the tutorials I read! I tried to amplify it but when I did, the background noise (including random static, probably because I was just using my phone to record) was very loud. I’ll give it another shot!

let me see what the equivalent steps would be in Audacity when i get the chance, and i’ll get back to you later…

do you have an original version of your sound file with the extra noise in it?

I would like to hear a longer audio to call it overprocessed, it sounds exactly as owls I heard, maybe original file was better, but not sure.

1 Like

Yeah, if the signal to noise ratio is not great, it can sound better to amplify by less than the tutorial suggests (e.g. -6 instead of -3 db). And if it still doesn’t sound great, best to just try and get closer to the bird and get a clearer recording. Recordings on a phone can be ok, but how close you are to the bird makes a big difference, and you’ll never get the same quality as with a directional microphone.

1 Like

can’t agree enough. Audio recordings are rarely improved with freeware editing and processing, even if you have pro software, you can only “polish a turd” so much. (industry term, calm down, admins.)

Proper microphone and preamp selection for your application will send you miles high than toiling for hours in Audacity or any DAW for that matter. Outside of a high-pass filter at about 250Hz, it’s best to keep your hands off.

A bad recording is like a bad photo, for example, if you shoot an image with extremely high ISO, there’s going to be hella “noise” (just like a phone mic or other not-purpose-built mics for full bandwidth recording) and there’s very little you can do (without professional software like PS or Topaz Labs stuff) that will “fix” the image without making it appear very unnatural, usually blurry or the like, the same applies for audio.

What you put in on the recording is the same thing you’re getting a playback, processing like EQ will only “mask” what you’re trying to eliminate (unless it’s a GML) it will generate a massive phase shift when frequencies are added or subtracted at their extremes, it’s the nature of how EQ works.

The best way to improve your recordings, is to make more recordings and make them better.

I think it’s very foolish to develop the habit of “fixing it in the mix”, it barely works, and it leads to a sub-par product and you’ll end up chasing your tail, especially if you’re trying eliminate noise, regardless of the source of it.

Sorry if I repeated myself or any of this seems unclear, I’m a bit under the weather today.

Happy to offer any advice or clarification for anything here anytime, I’ve been a professional audio engineer for longer than I’d like to admit :]

3 Likes

I can probably find it in my files, yes!

Thanks! I tried to keep the original owl sound as I heard it and filter out the noise, but I understand what people are saying. I definitely processed it quite a bit since there was so much “static” type noise that I tried to get rid of, I think that’s what people mean by overprocessed.

1 Like

Thanks. I’ll try amplifying less! One of my issue is that the noise is so loud, any little motion with my phone, the wind, etc. gets picked up and there’s a lot of a “static” sort of noise. I don’t mind if it’s not professional-grade, just good enough for an ID :) I’m not a serious enough bird recorder (yet?) to invest in a microphone!

1 Like

I use the editing protocol as suggested by eBird. It is similar to what you described here. I use Ocenaudio, but you can do the same thing with Audacity. https://support.ebird.org/en/support/solutions/articles/48001064341-audio-preparation-and-upload-guidelines

2 Likes