A mate of mine pulled out an old cassette of a recording from the late 70's that he played on, and ripped it to digital, handed it to me and said "if you feel like a challenge, see if you can clean this up." I do like a challenge.
The songs were really well done - great old Gospel hymns sung by a talented dude, backed with piano/guitar/bass/drums and a Gospel choir - I could hear through the bad quality that they were pros. But yikes, how do you recover that? Armed with my ability to use an AI model to do isolation, I set about trying to separate the instruments and vocals. Clearly the model was not made for such recordings, it couldn't separate out vocals/bass/drums/other, but it did a pretty reasonable job of separating vocals and other. I then had a chat to the Chatbots about what is going on a learnt a few new things.
My main issues were the usual hiss, boxiness and lack of sparkle, but the two other things that bothered me was the recording drifting from left pan to right pan, and the times the recording seemed to be out of tune. Some other minor things like vocals too loud at times and the drums were so mashed into the mix they were barely discernible.
So far, I have used the following approach and the initial results are quite satisfying:
- Separate vocals from everything else. The lead vocalist, not a problem, but the choir sometimes appeared in the lead vocals and sometimes in the other. Kinda makes sense, it was a dense choir, it wasn't too bad I found I could volume envelope my way around most of it.
- Collapse both tracks to mono, then re-expand them with:
- a hard left/right two tap delay which gives a mud-free perceived stereo effect. Depending on the difference in delays you can tweak how much expansion you want. My ears can detect even a 1ms left - 2ms right difference, it's subtle but it is there. At 5ms - 10ms it is quite clear. >50ms is starting to sound cavernous.
- Magic 7 is my favourite reverb at the moment which adds spatiality, but also adds muddiness for me, so I dial in a room sound and then adjust the wetness.
- EQ - lows and mids: I found the bass was booming annoyingly and the low mids were too prominent. I used a normal EQ to get this more like what I wanted to hear. I wanted to do a multiband EQ compressor but I find them hard to use - I need to practise more - so where there were sudden thumps I used a volume envelope to contain them. It's brutal, but if you don't push it too hard I don't hear it "everything suddenly ducked". Hmm, I really do need to practice using EQ compression.
- EQ - highs: Cassette seems to rob us of sparkle. When I tried to push it back in I was just increasing the hiss. So I used a little bit of EQ in the areas that didn't sound like his, and a saturator to create more high frequency overtones to "artificially" add more sparkle back. This seemed to work okay - what else can you do?
- Compression and limiting to help smooth out the dropouts and dips. And this might just be me, but I like smoothness and consistency in what I listen to. I like swells but I don't like to be jarred. Dynamics in music is great, maybe my ears are already too shot, but I like only subtle dynamics, I prefer a consistent balance. I don't want everything to be the same perceived loudness the entire time...well, maybe I do a little :-)
I noticed while trying to sweep the EQ listening for what annoyed me that "it seems to ring everywhere!" When I chatted to the bots about it, it's all artifacts from tape. Head misalignment, tape speed inaccuracies and just the nature of tape as a data storage medium. I found a compromise...these recordings are never going to sound crystal like they were recorded with modern gear, my main aim was to remove the panning warble, to balance the EQ, some re-producing/mixing to smooth out the dynamics, and I had to accept pitch drifts and out-of-tune sounds.
"Old school analogue vibe."
No comments:
Post a Comment