How to eliminate tinny echo in re-pitched voice recording?

[Larry343890152lwr]

Hope this forum allows the attached mp3. If not, will try to post it on a file locker..

I want to lower the pitch of a computer generated male voice - to make it deeper sounding and looking for a fix to the metallic like echo that doesnt change the volume.  MP3 quality is fine for this purpose.

Ran original file through Effects/Time and Pitch/Stretch and Pitch Process, and changed stretch to 90% and Pitch Shift to -3 Semitones. Don't have the latest Audition, hopefully your fix will work for older versions...

Played around with many effects, and watched a vid or two but not finding the answer...

[SteveG_AudioMasters_]

Have you tried Adobe's speech enhancer ? It certainly improves your sample, although it doesn't get rid of the ringing entirely. And as usual with these sort of issues, I don't think you will manage that completely because it's a dynamic artifact, and 'normal' effects have trouble with those. The speech enhancer works better because it uses AI to analyse what is actually speech, and reconstructs that, leaving everything else for the adjustment slider to control the level of. This effectively controls the  original artifacts, since there isn't anything else in the background to get rid of. RX has similar processing, but to be frank, the Adobe on-line version is currently rather better. So it might be able to help...

[Larry343890152lwr]

Thats much better, and a nice resource. Thanks Steve!

I should also ask - because Im only a self taught Audition user, and its unlikely I've chosen the best method, and want to learn a better way: What would be the ideal way to "lower" a voice?

Attached is an original file from the same text-to-speech website (freetools.textmagic.com) - maybe I could get a lesson?

The first thing I notice bringing files from this site into Audition is that the volume seems very low. I paste it into a new file to make it stereo, then raise the gain 6DB. Applying Stretch and Pitch afterwards to lower the voice really brings out the unwanted tinny, metallic echo.

Theres probably a better way to do this?

[Tony Zed]

This is about as good of a result as you're going to get out of adobe audition without any additional third party plugins. Even most third party plugins won't give you a flawless shift to the extent that you wouldn't be able to tell if it was shifted or just someone with an authentically deep voice. The human voice is one of those things we are so uniquely familiar with that the slightest unnatural variation triggers our brands to a phenomenon you've probably heard said in passing conversation, especially during this period in time when AI is become so prevelant. "The Uncanny Valley". Shifting the sound of most any sound effect or instrument or really anything else can be done fairly easy to the point that most people could tell there was maniuplation. But vocal pitching is generally used styalistically in some way, rarely intended to be a convincing re-creation. If you use Waves Vocal Bender or Soundtoys Alterboy, or even melodyne or autotune's manual pitch adjustments, you'll still hear an unnatural warble to it. With all that being said Adobe's pitch adjusting software is quite honestly some of the best I've heard to come as stock plugins in a some mixing application.

My best advice is to make sure and keep a back of copy because this will be destructive editing, and to apply the pitch adjustments inside of the waveform editor and render it to the track.

Pitch shitfting tends to be pretty intensive, and most often, the WORSE sound results you end up hearing is from trying to use it as a live adjustment. It cannot reach it's full potential of sound quality as an actively calulating instance. Find the settings that get you closest to the result you want, make sure it's the highest quality setting, and render it to the track and save so that in is baked in. That baked in version is gonna be the best sound quality you can get with what you have to work with.

Also, don't hesitate to try and record FOR your needs. The quality starts at the performance, not the post production. So if you want a quality deep pitch, start by trying to go as deep as you can in the original recording, and that will set you in the right direction.

[SteveG_AudioMasters_ · Accepted solution]

Most of that doesn't really apply to using a computer-generated voice - you don't exactly have a lot of choice about what you get. But there is one thing in there that you should be doing anyway, regardless of what you are editing, and it's something we mention frequently. Editing Rule 1: Always edit a copy, not the original.

As for the best way to go about the process - well, you will have to experiment. Time shifting generally brings out the metallic ringing, and this is inherent in the method that has to be used to achieve it, which is basically to slice up all of the audio and duplicate some of the slices. Or remove some of the slices, depending on which way you're going (this is a very simplified explanation of what's actually a pretty complicated process).

You may find, for instance, that you get better results if you do the process more than once, and only shift a small amount at each pass. And treat stretch and pitch separately. It's always worth looking at the advanced controls as well - generally it works better in single voice mode, with less ringing. But no two sources behave the same; experimenting is the way to go. Just note what works best for you, and you can use the settings again for the same voice easily if you save them as a custom preset.

[Tony Zed]

"Time shifting generally brings out the metallic ringing, and this is inherent in the method that has to be used to achieve it"

LOL, I think I was pretty much just attempting to say what you essentially said more concisely. 

I am intrigued at the idea of pitch shifting in multiple passes. Similar to the concept of serial compression, let each effect do it's own lifting gently instead of pushing one pass too hard.

I've found that to be true in many elements of audio mixing. When I have do pitch correction on vocals (this next part I know you already know Steve, but I'll still say it for the benefit of whoever reads this) Which is very different than outright pitch shifting, I have found time and time again that the most natural sound I could get was doing very gentle pitch correction in Wavestune (same thing as autotune, I personally like it better. Leslie Brathwaite loves it too so I'm in good company), and then doing a very gentle pass in melodyne, and doing those multiple passes always sounds better than trying to do it all in one or the other. So you definitely are onto something with taking that approach.

[Larry343890152lwr]

Thanks 10th and Steve for the explanations and insight. I'll try making changes in multiple increments to see if that helps.