Un-muffle music recording

Audacity 3 Help Forum Windows
,
1

Hello, I am new to Audacity, and had come across forums of using this to un-muffle music that was forwarded by a friend of mine that was found on an old forum. The OP managed to record a song played by simpy sticking the phone speaker to the wall, which in this case was very close to one of the speakers of a discotheque next door. However, I am unable to find the correct EQ settings to boost the music, so far all I managed to do was simply increase the volume that sounds like it will blow your ears off, but does not have the required information to allow smooth playback so that music recognition apps can identify the song.

I am using Audacity 3.7.3 and you can find the attached audio file for reference. If anyone can provide steps for general scenarios, or even identify the song, it will be a great help, thanks!

P.S.: I recently took to audio analysis coz of this project, but so far

I am simply floundering. Any knowledge or even self help/references will be greatly appreciated.

2

Its never going to be HiFi, but multiple applications of Filter EQ, and a Limiter, produced this

1 Like
4

Thanks for this try!

Though, I thought increasing the highs while keeping the lows would be enough to get the info of the song. So do music identifier softwares work by looking only at the highs, or is it something to do with the rhythm that the software catches in all frequencies?

5

Quote from Chat GPT when asked “how does music recognition software work?”
" Feature Extraction

The app breaks that audio into a spectrogram (a visual representation of the sound, showing frequency vs. time). It looks for strong features like:

  • Peaks in different frequency bands
  • Moments where frequencies change rapidly
  • Patterns that stand out (kind of like how our brain notices a catchy hook)

This becomes a “fingerprint” of the audio.

:dna: 3. Audio Fingerprinting

Each song in the database also has its own fingerprint, created in advance using the same method. These fingerprints are usually:

  • Small and efficient to store
  • Resistant to noise, background sounds, or distortion (so it still works in a noisy bar!)

A common technique is called constellation maps, where the app connects frequency peaks over time—kind of like drawing star maps of the song.

:brain: 4. Database Matching

The app compares the audio fingerprint it just recorded with millions of fingerprints in its cloud-based database. It looks for:

  • Exact or near-exact matches in the patterns
  • Time offsets (to match where in the song you started playing it)

Once it finds a match, it pulls up the corresponding song metadata: title, artist, album, etc."

1 Like
6

So to sort of “enhace” the fingerprinting of any audio file (music files being my focus of study), what concepts should I study as an amateur (sound engineer) so that I can understand how to bring clarity to dirty/noisy audio? For example, I came across another similar post where an engineer (I am guessing) provided simple recommendations to modify the characteristics of the audio file, which can be found here: EQ Recommendations

Should I start by simply playing around and understanding what each option does, or are there any resources that give a basic understanding of what levers to pull to enhance sound?

7

BTW, thanks for uploading this file, I managed to find that the song was The 1975 - Somebody Else, it took a while by Aha! Music.

1 Like