Author Options:

How can I create audio palindromes? Answered

There are plenty of guides to writing phrases that are read the same backwards and forwards. But how can I create some phrases that sound the same when the recording is digitally reversed?


Oh. At first, when I read the headline, I thought you were just talking about palindromic music. I've actually written an instrumental IDM song I titled "Symmetry". I never finished it, but it was interesting designing a groove specifically so it could be reversed. As it pertains to speech, this is a really tricky topic. There are several immediate problems. While a written palindrome is fairly easy to read and identify, speech is much more complex due to the fact that many phonemes vary from start to finish. Add to that the fact that accents (even British R.P. and American Midwest, which to their respective dialects are considered neutral and standardized) will change the outcome of the slides from one phoneme to the next, and you start to realize that there are a myriad of variables to account for. Another problem is that many phonemes are virtually impossible to pronounce correctly when reversed. Even simple ones. Try saying this and reversing it: Muck oitch no-w, nuss loh kalb, neeair eth ray woish owdna. Muck oitch no-w, nuss loh kalb. This is the chorus to Soundgarden's "Black Hole Sun" (which is a better song backwards than forwards - the chord progression and melody is phenomenally gorgeous backwards). Say this, then play it backwards, and it still doesn't sound convincing. Another prime example is Radiohead's "Like Spinning Plates" from their album "Amnesiac". Thom Yorke learned how to sing the lyrics backwards so he could record them and reverse the track to lend a creepy vibe. He did a fantastic job, but it's still barely intelligible at certain points. If you come up with a solution to these problems let me know; I've been dying for an answer to this for a long time, and I'm biting my lip in anticipation. I love playing with reversed stuff like this, and I would enjoy hearing some well spoken (or well sung) palindromes.

I'd neverheard the term either. Interesting Wikipedia article HERE .
I think audio palindromes would take a lot of work . . .

1) Create a "phonetic dictionary" - A list of letters and letter groups and their possible pronounciations (phonemes).
2) Record each phoneme and reverse it.
3) Listen to each reversed phoneme backwards and cross-reference to its nearest forward equivalent.
4) Build a meaningful palindrome with the first half from the 'forward' list and the second half from the 'backward' list of phonemes and . . . Voila!

It should be possible to write software to help with the phoneme comparisons and final stage. One thing about backwards speech is that the person is much more likely to 'get it' if they know what the message is meant to be in advance - for instance the backwards 'devil worship' phrases in heavy metal tracks and the 'messages from the dead' recorded in empty rooms.

I think this would be a really interesting project.

Here's a simple one I made up which I think should work. It's the referee's announcement of why one of the contestants in the duck-fight was penalised . . . "Pintail late nip".
There's a plethora of diacritical marks which indicate the exact pronunciation of a letter or word, but I've broken this down as a simple alphabetic equivalent, so the phrase is "Pi nn t ay ll ll ay t nn ip", which is phonetically palindromic..

Unfortunately, I think that "plethora of diacritical marks" was necessary. I couldn't make sense of the reversed version of "late nip" done with my usual pronunciation. (Isn't the phonetic transcription actually something more like p-h-ih-in-t-h-eh-ih-uhl l-eh-ih-t n-ih-p-' ?) Do you know how to write transcriptions in IPA? I don't know all those symbols, but could probably figure it out by examining the chart and audio examples.

Yes, there are extremely extensive phonetic dictionaries available but I think that using these to the full would make this project overly complicated and a smaller set of close approximations would be sufficient because of the brain's ability of making sense of distorted or incomplete data.

I wasn't suggesting speech recognition software. What I was thinking of was a program to easily select and play a single or sequence of phonemes, and then it's reversed equivalent. The listener would be the judge of 'quality'.

The 'pintail' example is one I made up whilst driving and haven't tested in any way. It 'feels' like it should work from the mouth-shapes involved, but I haven't tried recording it.

To me, there is a plethora of diacritical marks, mainly because I have never spent the time to sit down and try to understand them. I wasn't familiar with the IPA, but have now read the Wikipedia (English) IPA article. Although even with that degree of deconstruction of a word, a Geordie and 'Zummerzet' yokel would come away with totally different pronunciations from reading the example words. (If you're not from the UK, the example above will mean nothing to you. :¬)

This is one of those projects where you could consider the theoretical aspects endlessly, but the real progress would be to set up a test rig and try it.

Linguists already have dictionaries of words and phrases and their phonemes (some even with the exact allophones), so that part's done. Doing step 3 on software sounds like a good idea, though I don't know if any of the free speech recognition software works well enough. Step 4 sounds handy too. I'd appreciate any more ideas you have.


9 years ago

Like backmasking? I have NO clue.

I had to look up the term. Yes, I meant full backmasking and not just phonetic reversal.