Untangling doubled (orignal) recording

I have a challenging problem. I have transfered hundreds of old voice tapes using Audacity over the last 20 years, fixing many issues in the process but this is a tough one.

I worked in radio in the 1980s and i have many cassette tapes of my radio shows which i have been recording into mp3s. Some of these are ‘aircheck’ tapes – that is, the tape deck only recorded when the mic was hot, so it’s just voice (it was a kind of CV or resume for radio people). For some reason that i no longer remember, the aircheck recorders did not have an erase head, you had to manually erase the tape before use.

I recently found a very very important, long lost, 5 minute long section of an aircheck i had transferred to mp3 using Audacity 2.06 (on Win10). However on this recording a SECOND disc jockey later in the day recorded an aircheck on top of this one. The two voices are doubled. To be clear this problem was on the orignal cassette and not in the recording process in Audacity – they are not seperate tracks but fully stepped on.

In the past i have succesfully removed transmitter buzz or tape deck hum with Audacity by taking a sample of the sound by itself, but this is two voices talking (changing pitch), not a consistant hum. I don’t think i can envelope it like with a specific instrument because the voices are a very similar register.

I would really like to disintangle these voices so i can preserve the important section (i would probably discard the second voice). I suspect it is physically impossible. Would anyone have any suggestions?

Allegedly it is possible using AI … https://youtu.be/V5Mszp3-3C0 (male - female overlap).

( Not possible using Audacity’s native tools. Isolating all vocal content from noise/music is possible using OpenVINO AI plugin in Audacity, but it does not separate overlapping speakers ).

Is there music in the performance(s)?

Re-record the show using two actors in a quiet, echo-free room and note that this is a transcribed performance from the original tapes. I assume you can make out the words, you just can’t enjoy the show.

It doesn’t have to be magic. If you have a quiet environment, you can do this on your phone.

I recorded an audiobook-quality voice clip with that. Lossless Voice Memo on an iPhone. The cable is for power. Laying on the desk is intentional. That’s Pressure Zone Recording and it increases the quality of your performance with no further effort. The microphones are on the bottom pointed to me. The desk has to be quiet in addition to the room.

I don’t think you’re going to get software or an app to unscramble that.

I can think of insane ways to fix this. Transcribe individual words from the healthy performance and put them together into a “patched” version of the show. Can’t take you any more than six months.

Koz

If you get something to work, post back. This is a forum. Users helping each other.

Koz