What techniques would I use for improve this distortion.. ?

I don’t hold out much hope, but I’ve promised the boss I’ll ask:

We have a conference call recording that was done via Skype connecting to the conference call. Half way through, something happened to the conenction, drop in bitrate/low bandwidth would be my guess, and all the audio went like it does when youre on a mobile phone and only have 1 bar of signal - voices are robotic, drop outs, echo/reverb and words streeeeetch oooouuut/pops etc

My opinion is there isn’t a lot that can be done to salvage this because, let’s face it - if there was, then the clever beards at Skype would have built the salvation into the audio codec to make their product more killer market leading at low audio bitrates.

I’ve attached a small sample - is there anything that can be done, filters-wise to achieve even a 5% improvement in intelligibility of these voices?

You’re right, there’s not much that can be done with this. however, going forward I think we can help. How are you recording the Skype? Just having the far side much lower volume than the near usually means you’re not using Skype-compatible capture software. You can dig yourself a hole very rapidly by doing a Skype capture wrong.

And no, I don’t think it’s a data transmission problem because the near-side sounds bad, too. There’s no good reason for a transmission error to destroy the near-side voice. I was a party to a conference capture system that used a “dummy” conference participant whose job was to just hang on the data line to record and not talk. In that specific case, it’s possible to get all sides of the conference to sound funny in a network crash.

You can get a bad Skype capture system to completely screw up echo cancellation and that does affect both sides.

If your business depends on good clear recordings and you’re on Windows machines, Pamela is highly recommended. The two top licenses, business and professional will give you split, WAV recordings so you can process or filter each side of the conversation and then combine them into one clear “show” in post production.


There are other free products, but you are warned to stay away from software that insists on using lower quality MP3 files and that will not give you a split file.



Funny you should mention it, here’s the setup:

[conference line] -----skype group call A----- [one of our staff] -----skype group call A----- [recording device]

This is the audio part of it - there’s a conference call that multiple participants are on, giving opinions on research matters. They also use an interactive presentation. One of our staff joins the conference using Skype (voip → pstn) and they then also skype in another PC running on a cloud server that is watching the presentation. This way the audio stream from the conference reaches the recording device and is synced with what happens during the presentation. There isn’t a near and far party per-se, the relative loudnesses of the two callers are due to different phone (PSTN) configurations etc

It’s also why the recording turned poor - the staff member’s skype developed some issue and they reconnected to the conference. it seems that they then didn’t realise that the other outbound audio flow (by skype supporting its own mini conference between the main conf call and the recording device) to the recorder was poor quality… Hence all I have now is the audio that the recorder heard - the messed up one. In ordinary situations there is a backup audio recording taken by the conf call provider and also by the staff member using something like pamela…
Sadly those aren’t available due to a comedy of human errors, so this is the only audio streamI have :confused:

It’s all gone a bit Steven Hawking / “Speak-N-Spell”.

The signal has dropped in quality and they’ve [automatically] resorted to a very low-bit-rate codec, which is the Speak-N-Spell effect. And, like you said there’s drop-outs too. Nothing can be done to put back missing bits where there is drop-out or glitching.

You can make it all the same volume with Audacity , but that’s it …

I got nothin’.

I will note that we are not rescuing the voices from a very low-bit-rate codec. The very low-bit-rate codec is the rescue. Without that, the transmission would be unintelligible trash.

A friend was forced through circumstances to make a voice recording without listening to it. He watched his meters and dials carefully and made a clear, technically perfect sound file…of the wrong thing.