Separating Two Voices?

I know this may sound strange, but my mother was in charge of recording sermons at a church group…the tape she used was brand new, she had to cut the packaging off. When playing it back, at one point in the recording, another very low, very creepy demonic like voice takes over the micro-phone and you can hear some of the things it is saying, than a conversation is going on over-top of the speaker giving the sermon. Its a mono-track, and when in Spectrum mode you can clearly see the two separate voice frequencies…the speakers voice, and the demonic voice. I am trying to filter out the speakers voice to better hear this creepy voice in the background…is there a way to cut out the speakers voice based on the frequency or DB since it shows it as completely different? Can see it in my chart attached…the white part is the creepy voice that I want to isolate. I tried doing some research on it, using the low-pass filter and I just can’t seem to get it so that I can cut out the speakers voice, but still hear the creepy voice…anything I try makes it way to distorted to understand.

The white areas do not represent “voices” they represent the regions of greatest intensity at particular frequencies. The colours depend on the software being used and the settings in that software. There is no special significance to the colours other than different colours represent different intensities. In Audacity the colour white represents the greatest intensity so those frequencies will be the most dominant (most clearly heard) frequencies in the speaker’s voice.
http://en.wikipedia.org/wiki/Spectrogram
As you can see from this example of a male voice it is quite normal for multiple frequency bands to show up for a single voice due to the harmonic overtones in the voice:
http://en.wikipedia.org/wiki/File:Spectrogram-19thC.png
The background colouration is normal background noise in the recording.
There appears to be quite a lot of noise and some distortion in that recording but I don’t see any suggestion of anything demonic.

I see…I know nothing about this stuff…I figured the white area was the separate voice because it shows up when the other voice(s) kicks in. I’m not saying its definitely demonic…but it sure sounds creepy…the speakers voice is female, the ? voice(s) are male…but those voices have to be played back at 1.6x speed to better understand what they are saying because they are very low, and slow type of talk. How could this even be possible on a new tape? If it was given by someone else I would be a skeptic…but I trust my mother! I would also think it was somehow a re-used (but brand new in packaging) tape if the creepy voices in the background were not relevant to the speakers message…but they are…it seems they are mocking her as she is even praying. So, I appreciate the explanation of difference of colors…but is there an answer to my question? Could those separate voices be somehow isolated so I am not hearing them at the same time, talking over each other? That way I can than amplify and work with just the creepy voices…out of plain curiosity now of what they are talking about.

Separating voices from a mono recording is not possible because as you can see from the spectrograms even a normal voice spans frequencies of many octaves. If two voices are recorded onto the same track then the sound waves are inextricably mixed together.

It is not really our field but this article on EVP may shed some light on the phenomena that you describe. http://en.wikipedia.org/wiki/Electronic_voice_phenomenon

Ahhh, ok… I guess I just have to figure out how to better clarify the one voice over the other. I looked into the EVP articles a little bit…looks like it can be some pretty deep/complex/hoax type stuff! I didn’t get through it all, but seems the majority is EVP that people think they hear out of noise and pretty much force their mind to believe they are indeed hearing something. This is not like that… I included just a small sample of this audio clip so you can get the idea of what I am working with. There is no doubt about it, no " I think that might be a voice in the noise" its very obvious… at one point you can actually make out what they are saying, " You come over here…" and " I’ll be here all night" and of course the evil sounding laughter. Again, I would agree it could be interference from outside broadcast or something…but the parts I can make out, they are relevant to this speakers message, and if outside broadcast at the same time of recording…why the really low and slow tone is what I would question. I don’t know enough about sound to understand how at the same time, on a brand new tape, two voices can be recorded at different speeds/tones…only explanation to that would seem it was recorded over something else (but than had to be a used tape, which it wasn’t) and would have to be just coincidence that wherever the previous recording came from, it was in perfect timing to the current speakers message. As you can hear in the beginning, there is conversation going on in the background…but thats what I am unable to make out because I can’t figure out how to amplify those voices without amplifying hers…is that not possible? After a few seconds into it…you will hear the really clear parts that seems to over-power her voice in the microphone.

Definitely not EVP and definitely not demonic.

There is a conversation in the background between two woman that is at half speed.
Apply the “Change Speed” effect with a setting of 100% (double speed) and you will hear the voices much more clearly.

First woman: “Now, you come over here and you (get praying for him?). I’ll be here all night.”
Second woman: “You’ll be here all night?”
First woman: “well…”
… next part obscured …
Laughter (one person with deep laugh)
Laugher (group laughter)

As for how this happened I have no idea, but it sounds more like a conversation between two woment at a prayer meeting than something demonic. Do you recognise the voice of the first woman (when played at double speed)?

Haha…well you made that seem so simple (shows how much I don’t know). Being that they are women and talking about prayer makes sense all the more because it was an all womens meeting. ( Which originally made it seem on the demonic side when the voices sounded like men…and I guess I used the wrong playback speed option which cleared it up, but they still sounded like men). Guess the only strange thing left about it, is how it happened…and I’m not so much interested in that. Maybe somehow its playing through from the back side of the tape, if I go back and play the reverse side and hear that same conversation in normal speed…that would explain it! Thanks for your insight!

I use Windows 10 :neutral_face: and have Audacity 2.3.3 :smiley:
I have discovered that this is possible using noise reduction.
I have many old mono recording made in late 60s of my band. A couple had the vocals recorded on a separate track (on a domestic 4 track tape recorder) but the two voices, lead vocal and high harmony, were recorded via the one microphone. One mono track with two voices.
I found that by “sampling” the lead voice where it sings without harmony I could treat that as a “noise” sample. I then process the areas with harmony and remove the “noise” (my voice!) making the harmony voice louder but admittedly with the lead voice still there, albeit much reduced.
I was then able to put a) the noise track and b) the harmony track into my DAW (Reason 11) with usable stereo separation.
I have to admit that I didn’t actually use Audacity for this, but another non-free audio program, but I will try it in Audacity too. Why am I posting here? Because I was searching for other examples of the problem and solution (to see if I had discovered something new) and found your post.