One-minute test

All languages: post your test recording here. Help check audio files, provide editing services, and advertise for proof-listeners.
Post Reply
DyeffersonAz
Posts: 8
Joined: November 19th, 2018, 5:27 am

Post by DyeffersonAz » November 19th, 2018, 2:12 pm


Kitty
LibriVox Admin Team
Posts: 17749
Joined: March 28th, 2014, 5:57 am

Post by Kitty » November 20th, 2018, 5:44 am

all right, Dyefferson, I had a bit of time and listened to your test just now.

The reading is very fluent and you seem to have no problems with plosives, so that is already good.

As I already mentioned in the Contos phantasticos, there are a few things you need to correct in your settings though:

> the bitrate is wrong, we need 128 kbps constant (yours is variable). Do you use Audacity ? There you have to change the settings when you export to mp3. I can give you the step-by-step for Audacity, but other programs I am not familiar with.

> your DC bias is quite high. It means your recording waves are not "centered" enough around the middle line. This might distort the voice. In Audacity you can easily "normalize" and remove the DC bias.

> your background noise is quite high too. Unfortunately you didn't leave any ambient room noise at the end of the recording (we recommend 5 seconds). Usually there you can get a nice sample of the noise and clean it up a bit.

Here is what I would recommend:

1) re-record a short sample, you don't have to read it all, 10 seconds is enough.
2) after you finished speaking, let the microphone run for another 3 or 4 seconds, to get a noise sample without speaking
3) remove the DC bias
4) if the noise is very loud, try the noise-cleaning function of your editing program
5) when you export to mp3, make sure you have the bitrate set to 128 kbps constant
6) reupload and post the link in this same thread so I can have another listen

Thank you

Sonia

DyeffersonAz
Posts: 8
Joined: November 19th, 2018, 5:27 am

Post by DyeffersonAz » November 20th, 2018, 7:33 am

Kitty wrote:
November 20th, 2018, 5:44 am
all right, Dyefferson, I had a bit of time and listened to your test just now.

The reading is very fluent and you seem to have no problems with plosives, so that is already good.

As I already mentioned in the Contos phantasticos, there are a few things you need to correct in your settings though:

> the bitrate is wrong, we need 128 kbps constant (yours is variable). Do you use Audacity ? There you have to change the settings when you export to mp3. I can give you the step-by-step for Audacity, but other programs I am not familiar with.

> your DC bias is quite high. It means your recording waves are not "centered" enough around the middle line. This might distort the voice. In Audacity you can easily "normalize" and remove the DC bias.

> your background noise is quite high too. Unfortunately you didn't leave any ambient room noise at the end of the recording (we recommend 5 seconds). Usually there you can get a nice sample of the noise and clean it up a bit.

Here is what I would recommend:

1) re-record a short sample, you don't have to read it all, 10 seconds is enough.
2) after you finished speaking, let the microphone run for another 3 or 4 seconds, to get a noise sample without speaking
3) remove the DC bias
4) if the noise is very loud, try the noise-cleaning function of your editing program
5) when you export to mp3, make sure you have the bitrate set to 128 kbps constant
6) reupload and post the link in this same thread so I can have another listen

Thank you

Sonia
https://librivox.org/uploads/tests/teste_dyefferson_2.mp3

00:28

Here, I'm in another PC, and in another ambient as well, I tried what you said. What do I need to do?

Kitty
LibriVox Admin Team
Posts: 17749
Joined: March 28th, 2014, 5:57 am

Post by Kitty » November 20th, 2018, 12:43 pm

Well the good news is, you got the bitrate right now :thumbs:

but oh, there is really a LOT of background noise. It sounds like people are talking :hmm: For your official recordings I would suggest you find a quieter room ;)

Are you using Audacity ? Because it seems to me you don't know how to use the noise-cleaning functions and with Audacity I could give you a step-by-step.

What I now did to your last test was this:

1) I removed the very high DC bias. It was -185 ! (normal would be between -1, 0 or +1) so yours was really quite excessive
2) I reduced the volume which was over 94 dB, now it's down at about 89 dB
3) I ran two rounds of noise cleaning, because one was not enough to get rid of it all

Now here is the result, this would be already quite an acceptable test, even though in some parts where you speak (especially near the end) you can still hear people talking and this is noise we cannot delete without re-recording the sentence once again

Please have a listen to my cleaned version and you'll hear a noticeable difference: https://librivox.org/uploads/tests/teste_dyefferson_cleaned.mp3

This is what you need to be able to do with your editing program.

We can work on it together, but first you need to tell me what program you are using. And don't give up, we're getting there if you are persistent :)

Sonia

DyeffersonAz
Posts: 8
Joined: November 19th, 2018, 5:27 am

Post by DyeffersonAz » November 20th, 2018, 1:52 pm

Using Audacity 2.3.0 on the portable windows version

I have two computers, one that is Windows 7, and another that is Windows 10.

Thank you,
Dyefferson Azevedo

Kitty
LibriVox Admin Team
Posts: 17749
Joined: March 28th, 2014, 5:57 am

Post by Kitty » November 21st, 2018, 1:11 am

ok you have Audacity, this is great. I will give you the step-by-steps now, please use your last recording (not the one I edited, but yours) and do the following:

1) the DC bias:

> highlight your recording
> go to Effect
> go to Normalize
> tick Remove DC offset and press OK

2) the volume:

> highlight your recording
> go to Effect
> go to Amplify
> input -4 (don't forget the MINUS sign !) in the upper box (called "amplification dB")
> press ok

3) the background noise:

> select a small portion where you don't speak (usually at the end of your file, I took the sample from 0:27-0:28)
> go to Effect - Noise Reduction
> first select Get Noise Profile
> then highlight the entire recording
> again Effect - Noise Reduction
> now select OK

4) do this whole procedure of noise-cleaning TWICE, because you really have a lot of background noise

That should be all. As I said, then you still have some voices speaking inside your sentences, but for the test that doesn't matter. In the official recordings for a project though, the DPL would ask you to re-record those sentences.

Can you try these 3 steps and reupload once more please ?

Thank you

Sonia

Post Reply