Removing artifacts produced by software

O

o0Ampy0o

New member
Hello,

I did a search and decided to just create a new thread on this specific topic.

I am looking for advice on reducing or even eliminating (hopeful) the mechanical artifacts produced by audio software. An example is Audacity, a common and free application. The plugin included with Logic is similar and even superior but it also produces the unnatural sounding mechanical artifacts best described as that unnatural character getting farther away from a natural sounding voice and ever closer to a vocoder sound.

All of this type of software produces a mechanical artifact. It is easy to produce a robotic voice but it is difficult to significantly manipulate a voice and end up with a natural result.

I have purchased software that manipulates the voice in a wide range of ways from young to old, real and cartoony and then of course robotic which I could get with free software. I am trying to end up with as natural a voice as possible.

I imagine that the software company would include it in the product if it were possible so I assume that it does not exist. I have heard nothing from them in response to my inquiries. So I am here looking for tips to reduce/eliminate as much of this unnatural artifact as possible yet still make significant adjustments to the original voice. The more significant the greater the adverse consequences. I need to be able to change a voice from a young adult to elderly and child without it ending up sounding robotic.

Are there DAW tools or plugins that target this specific artifact?

Would something like SoundSoap help with this?

Thank you in advance.
 
Once an effect is exported on a 'natural' sounding recording, removing it is basically impossible.

I am not sure what your question is. Audio software does not generally produce any artifacts. Well, unless it wa recorded poorly with less than modern software.

You can't make a fish sound like a dog. Maybe a frog...

The original recording is what it is. Change it and then it is something different. There is no taking it back.


You really should post a sample of what you are talking about. What is it you describe as 'this specific artifact'?

'Artifact' to me is a digital anomaly due to an algorithm not doing what you want it to. Usually because either the software is not capable or the user is pushing the limits of said software.
 
Yes our definitions of audio artifact are the same.

The artifact exists pre-export. However there is software that treats a variety of unwanted sounds post-export.

I imagine any familiarity with the common audio artifact would be recognizable through my description of ever approaching a vocoder sound but a reference file would be useful. I will put one together and upload it.

The software is designed to manipulate the voice. Everything shy of a very mild degree of application produces artifacts so it is not pushing the software beyond its limits. It is probably pushing current technology beyond its limit at least at this price range. The software cost $300. Maybe something exists costing $3000 that can do it.
 
The voice of Ash after his head is knocked off in the film Alien is a good example of a voice showing similar artifacts. Think of his voice in a natural state. Compare it to the voice in the scene. That is what I am describing as sort of a vocoder-esque characteristic. It increases the more dramatically I alter a voice track with the software. This is how all of this type of software seems to work, at least that which I have been able to try. I would like to reduce if not eliminate this side effect. Otherwise the software achieves a convincing result.

 
You still haven't said what kind of processing it is (not the name of a program). 'Pitch correction, what?
We can use lots of these tools.. w/o theses problems ..unless and until they're pushed too far.
I guess second would be, why 'processing to remove the artifact -with more processing? Rather just focus on not getting into trouble in the first place.
 
mixsit said:
You still haven't said what kind of processing it is (not the name of a program). 'Pitch correction, what?
We can use lots of these tools.. w/o theses problems ..unless and until they're pushed too far.
I guess second would be, why 'processing to remove the artifact -with more processing? Rather just focus on not getting into trouble in the first place.
Click to expand...
Using any software known to you could you change a man's voice to a child's or woman's?

I need to be able to change a man's voice to a young adult or to an elderly person male and female without it ending up sounding robotic.

The software is a voice modeling processor. Something like pitch correction is a very small adjustment unless you are keys off.

I cannot do it using software at 5% of its capacity. This software changes multiple facets resulting in many characteristics typical of the target. For instance a child's voice is not just a difference in pitch. Among many things it has its own characteristic rhythm.
 
Last edited:
As has been said, once you've introduced and saved an effect, any artefacts will be virtually impossible to remove. In fact, forget the "virtually".

A few hints:

First, when you record the voice, get it as close to what you want at the time of recording. If you want a baby, do your best baby imitation. If you want an elderly woman, do you best imitation.

Second, always start small on the application of effects. Often, two or three "light" applications of an effect will produce fewer artefacts than trying to do them in one big pass.

Third, make good use of "Save As" to keep your earlier versions intact so, when you get to one you don't like, you can step back a version or two and try something else.

Finally, don't expect miracles from software (or even hardware) effects. They can only do so much before you start hearing problems.
 
Bobbsy said:
As has been said, once you've introduced and saved an effect, any artefacts will be virtually impossible to remove. In fact, forget the "virtually".

A few hints:

First, when you record the voice, get it as close to what you want at the time of recording. If you want a baby, do your best baby imitation. If you want an elderly woman, do you best imitation.

Second, always start small on the application of effects. Often, two or three "light" applications of an effect will produce fewer artefacts than trying to do them in one big pass.

Third, make good use of "Save As" to keep your earlier versions intact so, when you get to one you don't like, you can step back a version or two and try something else.

Finally, don't expect miracles from software (or even hardware) effects. They can only do so much before you start hearing problems.
Click to expand...

I have already said I am working with a native file not an exported file.

If I could make my voice sound like a child's or woman's I would not need software. There is a small contribution I can make by altering the cadence in my speech but that is only useful if the software does the other 99%.

It takes considerable muscle from software to change a man's voice into a child's or woman's. The software does exactly what I am attempting. I am not doing anything that I am not supposed to do with this software. But the consequence is that the result comes with the artifact.

This effect is used in film. Until recently I didn't know there was audio cleaning software costing $3000. There could be $3000 software that alters a voice without any artifacts.

I am looking for ways to reduce the artifact if there are any. Chances are some of it can be done using known techniques and less expensive tools. This software probably has many if not all of the dials to control various aspects of sound frequency. I know absolutely nothing about audio engineering. I was hoping someone might know that turning down or up a specific control could help.

I am going to experiment with the idea of applying the effect in multiple light applications. The way it works is that you designate a starting point (i.e. man, woman, child, etc.) My male voice is farther from a child's. A woman's voice is already closer to a child's. If I can maintain a clean source file to apply lighter treatments I might be able to attain my goal taking this alternate route. However, I foresee this adding an additional problem in that any lingering artifact will be reprocessed rendering the final result more mechanical than that achieved in a single application of the effect.

The Save As work-in-process is a good general approach I use with everything from desktop publishing to visual design.

I am doing voices for all of the characters in an animated cartoon. I need to create recipes achieving different voices for various characters not just one. Each character will need its own formula. Consistency becomes a factor. I am better off with everyone having the same artifact than each having varying degrees of it. I have come up with a plot to explain the artifact if a solution eliminating it cannot be found. I thought of it before I made the decision to invest in this software. But I would prefer that the artifact be eliminated so I have the freedom to compose stories outside the constraints of this explanation for the artifact.
 
Last edited:
You mentioned Audacity and Logic. Are these the software programs you are using to manipulate your vocal tracks? Or are you using something else? Tell us what you are using.

You can look at Melodyne. It has a limited ability to alter voices to change range and octaves. It allows editing of formants which give a voice gender qualities. I don't completely understand how it works, but formants are the part of phonics that we can't alter with pitch. Pitch resonance formed by sinus cavities, size of the throat, head, etc... Something like that.
 
Why are you picking on the DAW? It could be the mic or the driver you are using with the mic, or your sound card.

If you want to get as close as possible to "natural", I assume that you'll need all the different components to be compatible and the end to end product built to record "faithfully". mics are built for different purposes, and have a response curve to suit, which may not be faithful to the original sound.

Then the mic driver can totally change the sound. I tested this only a couple of days ago. When I connect my mic, Audacity chooses which driver to load for the mic, and it doesn't always choose the same one. The difference in the recording is huge.

I think it decided to load the correct mic driver, here...



And here, it loaded another driver. I was using exactly the same equipment. And it sounds more robotic, to me.



Then you've got the sound card, both its recording and playback characteristics. Then you have to make sure that the artifact is not caused in the playback, by the sound card or speakers/headphones.

In my case, since everything was fixed, I am sure it must be the mic driver that Audacity chooses to load.
 
kickingtone said:
Why are you picking on the DAW? It could be the mic or the driver you are using with the mic, or your sound card.

If you want to get as close as possible to "natural", I assume that you'll need all the different components to be compatible and the end to end product built to record "faithfully". mics are built for different purposes, and have a response curve to suit, which may not be faithful to the original sound.

Then the mic driver can totally change the sound. I tested this only a couple of days ago. When I connect my mic, Audacity chooses which driver to load for the mic, and it doesn't always choose the same one. The difference in the recording is huge.

I think it decided to load the correct mic driver, here...



And here, it loaded another driver. I was using exactly the same equipment. And it sounds more robotic, to me.



Then you've got the sound card, both its recording and playback characteristics. Then you have to make sure that the artifact is not caused in the playback, by the sound card or speakers/headphones.

In my case, since everything was fixed, I am sure it must be the mic driver that Audacity chooses to load.
Click to expand...

Dood... Not the point man and not even sure what a mic driver is referring to. USB mics? Not what the OP is talking about.
 
jimmys69 said:
Dood... Not the point man and not even sure what a mic driver is referring to. USB mics? Not what the OP is talking about.
Click to expand...

Well, I am curious how you managed to figure that one out from the OP.

My point is, how do you know that it is the DAW, rather than the hardware or some other software that is producing the unwanted effect? Are you saying that there are no mics, for example, that sound robotic or unnatural?
 
kickingtone said:
Well, I am curious how you managed to figure that one out from the OP.

My point is, how do you know that it is the DAW, rather than the hardware or some other software that is producing the unwanted effect? Are you saying that there are no mics, for example, that sound robotic or unnatural?
Click to expand...
Well, we're finally getting the OP to at least start getting specific :rolleyes:
But it seems clear the trouble is in doing massive mods to voice formant alteration -where he/she also added that 'Alien thing's swirling' .. Likely not talking 'a mic or hardware problem.
 
kickingtone said:
Why are you picking on the DAW? It could be the mic or the driver you are using with the mic, or your sound card.

If you want to get as close as possible to "natural", I assume that you'll need all the different components to be compatible and the end to end product built to record "faithfully". mics are built for different purposes, and have a response curve to suit, which may not be faithful to the original sound.

Then the mic driver can totally change the sound. I tested this only a couple of days ago. When I connect my mic, Audacity chooses which driver to load for the mic, and it doesn't always choose the same one. The difference in the recording is huge.

I think it decided to load the correct mic driver, here...



And here, it loaded another driver. I was using exactly the same equipment. And it sounds more robotic, to me.



Then you've got the sound card, both its recording and playback characteristics. Then you have to make sure that the artifact is not caused in the playback, by the sound card or speakers/headphones.

In my case, since everything was fixed, I am sure it must be the mic driver that Audacity chooses to load.
Click to expand...
By the way, a 'bump for your post -very interesting those sound clips and a good contribution. :)
 
mixsit said:
Well, we're finally getting the OP to at least start getting specific :rolleyes:
But it seems clear the trouble is in doing massive mods to voice formant alteration -where he/she also added that 'Alien thing's swirling' .. Likely not talking 'a mic or hardware problem.
Click to expand...

The impression I get from the OP is that he has a vanilla setup with no intentional mods, but he is still getting a distortion of the sound by something. And that distortion is messing with the voice alteration software he wants to use.

^
And thanks for the latest :-)
 
Given your sound clips- examples, yeah I can now see where that could be in play. Shoot, my wold has been RME', Sonar and stuff, I had know friken idea :>)
 
kickingtone said:
Why are you picking on the DAW? It could be the mic or the driver you are using with the mic, or your sound card.
Click to expand...
Where have I or anyone else picked on a DAW? In my OP I just asked whether a DAW had adjustment tools that might help eliminate the artifact. That was the only time a DAW was addressed. Where do you get the idea that anyone has faulted a DAW?

Your sound samples are not relevant. Between your two samples there is nothing that really adds or detracts from the age of the vocalist. The largeness or smallness lends more to position in 3D sound space than anything. One sounds closer while the other more distant. The fullness of the sound, the main difference I detect between your sound samples, is not the problem.
 
mixsit said:
Well, we're finally getting the OP to at least start getting specific :rolleyes:
Click to expand...
One need only ask a specific question. I tried to provide information without writing so much no one would be interested in reading it. If you think of something missing just ask me a question.
 
Chili said:
You mentioned Audacity and Logic. Are these the software programs you are using to manipulate your vocal tracks? Or are you using something else? Tell us what you are using.

You can look at Melodyne. It has a limited ability to alter voices to change range and octaves. It allows editing of formants which give a voice gender qualities. I don't completely understand how it works, but formants are the part of phonics that we can't alter with pitch. Pitch resonance formed by sinus cavities, size of the throat, head, etc... Something like that.
Click to expand...
The software is Flux Ircam Trax V3. Melodyne appears to be mainly targeting pitch correction. It appears to do one thing well. It is not specifically designed to fulfill my objective, to alter a voice to change its gender and age. This software is. I mentioned Audacity because a lot of people are familiar with it and might have discovered the artifacts it typically generates when making significant alterations to the sound file. This software can do pitch correction as well as many other things but it is not refined to do it like Melodyne appears to be. Melodyne looks like they have focused a lot of energy on making the pitch correction process as efficient and simple as possible. I am sure it is nice and may be the best according to its users. I don't think it will get me in the ballpark of what I am aiming for. I might try the trial later but first I am trying to work with what I have.
 
You must log in or register to reply here.
Back
Top