Smooth out tempo and pitch for old cassette recordings

I have some cassettes recorded many years ago on a battery powered sony walkman. After transferring several to mp3 I have realised that they tend to speed up and increase in pitch- probably because the batteries were going flat when doing the original recording.

They are live band recordings and typically about 10 songs approx per cassette side. Each song will have a different tempo. Is there a simple (or if necessary complicated) way to straighten out the tempo and pitch so that they stay constant. I am using Cubase 10.5 pro.

Tempo detection might be useful. According to the manual:

Detecting the tempo of musical material is useful for the following cases:

  • If you want your audio or MIDI tracks to follow the tempo of freely recorded material.
  • If you want to adjust freely recorded material to the project tempo.

I haven’t used it myself so I can’t do anything more than pointing you in that direction.

But I thought all the kids were buying software to make their stuff sound that way on purpose?? You mean you want to fix it?? ; )

I think any tempo adjustment will leave the pitch wrong. I think you need to do tape style pitch adjustments until the tempo is constant. Maybe even put the audio in a sampler track and use pitch bend to fix it. Generating a tempo map can help show you where you need to make your adjustments.

This reminds me of a previous thread where the considered opinion was that this is a very hard ask, usually accomplished by specialist hardware/software.

I’ve had some success in a similar project by generating tempo maps. The algorithm elastique tape gave favourable results with minimum degradation, but my poorly archived R to R tapes were beyond redemption!

I second GlennO’s suggestion of pitch bend in a sampler track.

Thanks for the suggestions. Its a thankless task. I might just concentrate on the most obvious bad sections and leave the rest. Too busy fighting with Cubase on other things at the moment!

I may be wrong, but doesn’t Celemonys Capstan plugin correct wow & flutter etc. from tape recordings? Might be worth a look.

I’ve thought about this a lot as I have heaps of old 4 track recordings I aim to digitize when I get time.

https://www.plangentprocesses.com/ looks like the gold standard, but unfortunately it’s not accessible to we mere mortals…

I’m going to try capturing the cassettes at 192khz, hopefully allowing me to use the bias frequency as a timing reference. The yamaha mt120 4-track that I used has an erase bias frequency of 80khz (not sure what the record bias is - hopefully under 96khz!).

As far as using the bias tone - that’ll be an interesting experiment. Maybe isolate the tone with spectralayers, then pitch correct it somehow with melodyne, then work out if I can apply the same pitch correction curve to the audio…

Could be possible with Capstan. Isolate the bias tone with filters, let Capstan generate correction, then (from the Capstan manual:)

Apply Speed Curve to other files…
You can apply the correction of one file to any number of other files. Prerequisite
is that those other files share the same attributes (sample rate, bit depth and file
length) with your source file.
This feature comes in handy in a variety of situations:
• Surround mixes: Perform your analysis and edits on the L/R channels and apply
them to the Center and Surround channels.
• Multitracks: Perform your analysis and edits on a temp mix of all channels and
apply them afterwards to the individual channels. This way you can also control
which channels you want to include in the analysis. You can exclude challenging
channels like the vocal tracks to get a more robust wow&flutter detection.
• Experimental: Try steep filters to filter out sound that disturbs Capstan’s
analysis. > Create a temporary file just for analysis and editing and then apply the
Speed Curve to the original file.

I’m sure it’s possible with Capstan, it’s the very highly thought of product built to solve exactly these issues but it costs €3,790. Even a five day hire costs €199.

I expect that’s what brycem means when he says it’s not accessible to mere mortals.

Haha, no I actually meant the Plangent Processes - there’s no indication of price or whether the hardware and software is actually for sale on it’s own, kinda has an aura of hyper-exclusivity about it on the website (Neil Young, Bruce Springsteen, Grateful Dead have used it).

Actually, digging deeper, it sounds like it may just be a service where one sends tapes or high res digital tape captures to them. Seems they do a similar thing to what I suggested with using Capstan on an isolated bias tone using their own “Clarity” software. Here’s a description from the liner notes of a Grateful Dead release:

The first step in this process was to transfer the original analog master tapes to a digital format. This was done by using a Pacific Microsonics HDCD A/D converter at the highest possible digital resolution, 192KHz/24bit (for comparison, a standard CD is 44.1KHz/16bit).

The resultant audio files were then sent via hard drive to Plangent Processes in Nantucket, MA. Plangent Processes utilizes a novel proprietary digital signal processing system called Clarity. > The process analyzes the tape, searching for artifacts of the recording stream that happen to contain highly accurate timing information, which then reveal the mechanical imperfections of the original tape recorder’s performance as it was making the initial master recording. To obtain this timing information, the original transfers to digital actually had to be done with the master tapes running at half speed, 3.75 inches per second.

This information was then employed as a guide to gently but firmly correct the data of the digital transfer such that the sonic performance in terms of speed, pitch and tempo was indistinguishable from the board output. In this case the original 7.5 inches per second recordings contained significant amounts of wow and flutter, as would be routine even in the professional machine used to make these recordings. By undoing the wow and flutter we can now hear the performances in their original perfect pitch, and with steadiness of rhythm and overall clarity intact.​

Playing the tape at half speed is a great idea that hadn’t occured to me, that means I’ll be able capture bias tones up to 192khz on my cassettes that were recorded at 9.5 cm/second by playing them back at 4.75 cm/second :slight_smile:

So my plan going forwards - capture the tape playing at half speed into adc at 192 khz, Then inspect with spectral editor to find bias tones. Then filter out everything else, run through Capstan, then apply Capstan correction to unfiltered files. Then resample to get back to normal playback speed. I’m keen to see if this actually works, but my 4-track is in bits on my workbench awaiting a chance to install a brand new playback head :slight_smile:

And to clarify, I used a demo of Capstan a few years ago and had bad results - it locked onto tones in the actual audio and then tried to correct any vibrato or pitch modulation in the performances. Isolating a bias tone as timing reference should yield far better results!

Wow…and interesting.

I believe that it’s possible in Capstan to tell the program to lock to a predefined pitch (e.g. A=440Hz or whatever you decide) or to an average pitch In the specific project. There are a few parameters to play with and from what I’ve seen it’s a child’s play to achieve perfect results without getting artifacts from vibratos or pitch modulations. The price of the software though is otherworldly, sort of…

You could try a 30 day demo of this… looks rather interesting! https://www.youtube.com/watch?v=DznkILg6XOM

Heh, you beat me too it, had just seen this myself and remembered this thread. Does indeed look interesting if the demo is completely unlimited for the 30 days (I think they usually are with iZotope).

It’s pretty impressive, from what i’ve seen and heard so far, and A LOT cheaper than capstan, seems way more straight forward to use too.