annehaviland's test mp3 [OK]

All languages: post your test recording here. Help check audio files, provide editing services, and advertise for proof-listeners.
annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 16th, 2018, 4:32 pm

Hi, I'm a new volunteer reader to Librivox and I have recorded the test mp3 using Audacity and an Audio Technica carioid condenser mic. I plug the mic into a Focusrite 2i2 audio interface using an XLR cable. I'm looking forward to your advice about how I can tweak settings for a better sound. Here's the link: https://librivox.org/uploads/tests/test_annehaviland.mp3

My environment to eliminate echo and ambient noise is 1) I have a sound shield around my mic and 2) I created this portable "room" ...I have two 6' rolling clothing racks and I hang thick, soundproofing blankets on them to create walls 2 blankets thick from 6' above to the floor. I can move these "walls" out or in a little but not much, as my office is small. Anyway, with the blanket walls and curtains drawn I've managed to get rid of a lot of the echo I was first hearing. But your ideas and suggestions are welcome.

Thanks!

audiomike
Posts: 13
Joined: May 27th, 2016, 10:26 am
Location: Mid-Michigan

Post by audiomike » September 17th, 2018, 6:59 am

It's refreshing to see (and hear) that a new reader has put some effort into the quality of their recording on the first try. I realize that many people don't have your level of equipment, but it's the effort that counts. All the people out there that use noise cancelling headset mics or their built in laptop computer mic three feet away from them in their kitchen, need to read your post and listen to your recording to get an idea of what a good one sounds like. Your voice is clear and your noise floor is still close to -50db without any effects after normalization to -3db. As far as the "tech specs" go, just from listening, I think it needs to come up some in volume level. I'm sure one of the usual PLs will let you know all the techie stuff you need to do. They're better at doing that kind of thing than I am. Once again, good job!

annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 23rd, 2018, 7:38 am

Hello Audiomike,

Thanks so much for listening to my recording and giving your input. I'll look into improving my volume. I have the equipment but still don't know a lot about using it, so I'm not sure if the volume needs to be adjusted on my interface box or in my recording software (Audacity) but I'll do some research into it. On a related note, I narrated a short story I wrote that has a scene of high conflict - lots of yelling. When I played it back, I heard that my yelling voice was distorted. It seemed to sound digitized. I'll look into how to correct or prevent that. I narrate training at my day job and have been spoiled because that type of recording doesn't require a lot of know-how beyond basic. It'll be good for me to get out there and learn more! This forum is a good start. Thanks again!!!

annehaviland :D

annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 23rd, 2018, 7:46 am

PS: Audiomike, if you return to this thread, could you explain the significance of having a noise floor close to -50db without any effects after normalization to -3db? If I understand your input correctly, this characteristic is a good thing, but am not sure why. Again, thanks!

annehaviland

audiomike
Posts: 13
Joined: May 27th, 2016, 10:26 am
Location: Mid-Michigan

Post by audiomike » September 23rd, 2018, 10:28 am

I'm not quite sure how to explain all this so it makes sense but I'll give it a try. After all, you did say you wanted to learn more.
We'll start with how to handle the yelling. Without the ability to use a hardware compressor/limiter in the recording chain, you'll have to account for most of it manually during recording. What this means is that you have to figure out where to set the input gain so that the loudest portions don't clip during recording. You can use the halo rings on the 2i2 to know if you're clipping or not at the loudest points. The problem here is the wide dynamic range you wind up with. That's the difference between the quietest and loudest parts of the recording. Now, for music you want a wide dynamic range. For spoken word, however, you don't need that much to still sound good. So we need to squeeze the quietest and loudest parts closer together. In Audacity, after recording, you use limiting and/or compression to squeeze the dynamic range so that you can hear the quietest parts at a normal volume with the loudest parts only slightly louder. In an audio book you don't want to jolt the listener with sudden large bursts of volume. Getting this right takes some trial and error. The best way learn how to do it is to look for compression and limiting in Audacity on YouTube. It's way too much to try to explain here but I can answer specific questions if you have them.

Now for the noise floor. What I meant was that you were able to achieve a -50db noise floor without applying any effects in Audacity. Few people can accomplish this feat without using a considerable amount of Noise Reduction which makes their recording sound bad. To give you an idea of where you fall, a lot of Librivox recordings strain to reach -30db whereas on the other end Audible requires -60db.
And the explanation: Normalization is defined as: "the application of a constant amount of gain to an audio recording to bring the amplitude to a target level (the norm)". This usually means bringing the amplitude up to either -1db or -3db. I use Adobe Audition, not Audacity, and when I bring in a Librivox recording, the first thing I do is normalize it to -3db. Then I check the Noise Floor. That is simply the level of the noise, not the overall recording. When I see a raw file (one with no effects applied) that hits -50db, that's pretty special for a new reader. By the way, that -3db target is about right for Librivox.

I hope all this made some sense.

An unfortunate drawback to the Livrivox system is that readers are also required to be producers. Needless to say everyone eventually finds out that reading is a lot easier than producing. The PLs do the best they can to help but they're also working with limited knowledge of the audio production process. If you're willing to learn, you will be able to create a final recording you can be proud of and maybe help others down the line.

schrm
Posts: 785
Joined: February 10th, 2018, 11:02 am
Location: Austria

Post by schrm » September 23rd, 2018, 11:09 am

hi annehaviland and audiomike,

as i know, that your knowledge about audio is way above my abilty to understand, i did the pl for this .
so, just to throw it in:
:!: the average volume of the test file is 79,2db, we aim for 89db.
:!: the ending silence is too short, we aim for 5 seconds.
:!: at second 23,25, there is a clicking sound.
otherwise, the test recording is ok - as it was to be expected.

audiomike, please, can you help anne to meet our 1minute test criteria?

cheers,
schrm
reader/12275


deutsch
wunderwelten viewtopic.php?f=60&t=70793
sammlung kurzer prosa viewtopic.php?f=60&t=71431

english
violet viewtopic.php?f=2&t=70567

audiomike
Posts: 13
Joined: May 27th, 2016, 10:26 am
Location: Mid-Michigan

Post by audiomike » September 24th, 2018, 7:30 am

schrm, no problem. I'll give it a shot.

Anne, Let's get the easy ones out of the way first.

IMPORTANT: Never work on your original file. Save it, make a copy, and use the copy for edits. If you mess it up, make another copy and do it again.

The clicking sound is probably a mouse click or a switch on the mic. In either case you can just delete it or reduce the level. To reduce the level, select the click area, go to Effect/Amplify, put -9 in the top box and click OK. You should get into the habit of waiting for 5 to 10 seconds before and after recording without any background noises such as mouse, keyboard, and switch clicks. You can delete bits of the unneeded silence later to be within required specs.

For the needed silence at the end just select about 3 seconds, copy it (Edit/Copy), click out of the selection within the space at the end and paste it back in (Edit/Paste).

The volume part. This is the one you want to remember since you'll use it a lot.
Start by selecting your entire file (Select/All)
Got to Effect/Normalize
Make sure the Remove DC Offset box is checked
Enter -3 in the box and click OK (you'll want to use -3 all the time)
This should give you about 89.6db

Now for the audio lesson.
There are about 8 or 10 places in the recording where the peaks are way above the rest of the audio. When you Normalize (or amplify) Audacity is going to raise the level of the highest peak to your reach your target, -3 in our case. This leaves the quieter parts of your recording amplified but still too low. So let's fix it.
Select the entire file
Go to Effect/Limit
Make sure Hard Limit is selected in the top box
The rest of the boxes should be (from top down) 0, 0, -10, 10, No
Click OK. Notice that the waveform just squeezed down. The peaks have flattened but they are not clipped, just lowered to a preset level.

Now, because we just lowered the peaks, we need to Normalize it again to bring them back up to -3.
Follow the steps above in "The volume part" and that's it. You now have a recording that meets the target volume of 89db and has a more consistent level of voice and we did it without altering the dynamics of the recording. Listen to your original recording and the edited version. Leveling out those loud peaks makes it easier on the listener's ears.

Well, you should be thoroughly confused by now. Let me know if you have questions.

annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 24th, 2018, 4:50 pm

LoL only somewhat confused :D Thank you both very much for your direction, explanation, and advice. I copy/pasted your posts to Word so I'd have a cheat sheet while working. I never have edited on a copy, always the original; so I'm glad to get that word of warning. I'll make it a habit to work from a copy. I'm psyched to try out the editing. Thanks again!

moniaqua
Posts: 1152
Joined: April 11th, 2013, 4:48 am
Location: Somewhere in the south

Post by moniaqua » September 24th, 2018, 8:55 pm

Man, what a novel!
annehaviland wrote:
September 24th, 2018, 4:50 pm
only somewhat confused
Although audiomikes advice sounds interesting, for LibriVox it is totally sufficient to stick to schrms list, which might be less confusing :) So, basically I would
  • Go to Effects - Amplify and amplify 9.7 dB (that's the number Replay Gain gave me; you can also use Checker, see link in sig)
  • mark the silent end and go to Generate - Silence , click in the arrow besides the number in "duration", choose hh:mm:ss an type in 5. That adds instead of the marked part 5 seconds of silence. You could also just cut out the bump and copy a part of the silence that you reach the five seconds.
That's it. You might, to perfect it than, brush just a tiny little bit over the noise, but that's not really necessary.
annehaviland wrote:
September 24th, 2018, 4:50 pm
I never have edited on a copy,
Mostly I don't, either :oops: There's that wonderful "go back"-option :D But it can't harm to work on a copy; especially on longer texts. You never know...

audiomike
Posts: 13
Joined: May 27th, 2016, 10:26 am
Location: Mid-Michigan

Post by audiomike » September 25th, 2018, 7:13 am

Well, I seemed to have caused a bit of a stir again. My intention is simply to give good readers that have a desire to learn a bit of useful knowledge that can make them better. I fully realize that PLs are working within the system. I also realize that they are dealing (mostly) with people that have no idea in the world what they're doing. By no means does the reader have to do anything I say. I'm just throwing it out there. I guess I should have also included the easy way. Sorry about that.

That being said, I would like to rectify an error in the previous post. If, that is, anyone wants to try it. In Audacity, the Normalization effect needs to be at -6db, not -3db to provide the proper output level that will make Checker happy.

One more bit of advice and I won't bother you anymore. If you don't really, really, need to use Noise Reduction, don't do it. Eliminate the noise at its source, not later. Amplification, Limiting, and Normalization simply raise and lower volume levels. Effects like Noise Reduction affect the overall frequencies that make up the audio. That is why it's usually noticeable when someone uses it. It degrades the sound of the voice. Contrary to popular belief, there is no need to have absolute silence in the gaps. In fact, it's rather unpleasant to listen to. A bit of what is called "room tone" is actually necessary and helps provide consistency to the overall sound.

Once again, I'm sorry if I ruffled any feathers.

annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 29th, 2018, 3:08 pm

Hi audiomike, thanks! I edited (a copy, not the orig!) using the advice you gave before. I'll go back and try editing with -6 in Normalize instead of -3. Then compare to hear the difference. That's interesting about the noise reduction not being useful and that a little room noise is more comfortable to the listener. Noise reduction is my routine at work, so this experience with Librivox is breaking some old habits. Back soon with an edit mp3. :D

annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 29th, 2018, 3:14 pm

Thanks, Maniaqua! I'll try your edit steps, as well as those audiomike gave me. It will be interesting comparing to hear contrast in output. I'm having a really good time learning new editing tips from you guys. I really appreciate the time and attention you've given me.

annehaviland
Posts: 13
Joined: September 15th, 2018, 12:16 pm

Post by annehaviland » September 29th, 2018, 4:11 pm

Hi guys, I edited my test mp3 according to the advice you gave. It was a fun experiment to create a test-edit using each set of instructions. Of course when I used Checker, both edit files passed :)

This is the file edited with audiomike's process: Normalization at -6; Limit at 0,0,-10, 10, No; and Normalization again at -6; with the nonverbal end elongated by copy/pasting it until I had 5 seconds:
https://librivox.org/uploads/tests/test_annehaviland-edit-2.mp3

And this is the file edited with maniaqua's process: Effect: Amplify at 9.7 dB and 5 seconds of silence replacing the nonverbal end.
https://librivox.org/uploads/tests/test_annehaviland-edit-3.mp3

Thanks schrms, audiomike, and moniaqua! I'm really looking forward to reading -- and editing :D

schrm
Posts: 785
Joined: February 10th, 2018, 11:02 am
Location: Austria

Post by schrm » September 30th, 2018, 12:15 am



good morning!

(here it is :-) )

first: audiomike, thank you for tutorial! but, as i told you before: since lv is a work of volunteers, it was a little bit much to do at the beginning.
but i think,you yourself are on good way: reducing here a little bit, forcing there a little bit into the librivox-ways...
maybe, to volunteer at librivox (according to lv conditions) could suit your interests and aims? 8-)

second: either way, thank you monika, for putting this back to librivox-levels!

:9: third and last but not least: your testrecording is perfectly ok now, both recordings meet our technical criteria.
but one: please never change the title of your recordings, also when uploading another version. it eases the use of testrecordings, but makes things complicated in book-projects. the lv-software adds a date to the old contribution and the resulting audiobooks should have the titles "in order", therefore we need you to upload new versions with the "old" title as found in every first post of a book-project-thread!
now, since you edited your recording twice, i totally understand, why you did that and i thank you for that :-)

some remarks:
thank you for putting so much effort in this and doing it twice!
you didnt need to do that, but now i ask you:
which result do you like better?
you can also do a mix of both ways, since you are contributing recordings to your own style..
i prefer the fast to do way (see below, why i write this) and liked the third resulting recording better. your voice has more harmonics or bass, i think. but that is not up to me to decide (or to talk about equalizer and stuff, i never use)!

so, one more hint at the end of my result-post, because you made so much effort with this:
please, dont get frustrated by endless editing! we are just a volunteer-project, and we dont want to frustrate volunteers.
i have to add: i stopped to read for some weeks now, because i edit too much and too long every recording.
additionally, i have real life stress at the moments, there are builing sites around, sport events, and so on - and i began missing my hobby-work for lv badly, already.
so, the most important point for me is, at the moment, to contribute short sections, fast to edit, in pauses of noise around me, when im able to read.
to avoid this, i also reduced my levels of claims for my own recordings..
so, very up to date, i warn you, that real life can interfere, and has already with many many people.
one way to deal wth this is then: back to the basics, like we pl the 1 minute tests :-)


and now to the fun aspects, again:
did you have a look around and know already, what you want to read first :?:
most people prefer to begin with some short texts or poems - please have a look around and claim a section or two, which suits your interests or catches your eyes :D

cheers,
schrm
reader/12275


deutsch
wunderwelten viewtopic.php?f=60&t=70793
sammlung kurzer prosa viewtopic.php?f=60&t=71431

english
violet viewtopic.php?f=2&t=70567

moniaqua
Posts: 1152
Joined: April 11th, 2013, 4:48 am
Location: Somewhere in the south

Post by moniaqua » September 30th, 2018, 1:22 am

Good morning.
annehaviland wrote:
September 29th, 2018, 3:14 pm
Thanks, Maniaqua!
You're welcome :). Is there a certain reason you call me maniaqua? :hmm:

As schrm already has given you the ok, I don't need to to this :) Thank you for your patience with all of us.

I won't tell which one of your recordings I liked better but I am happy that you can choose out of two for yourself functioning workflows.
annehaviland wrote:
September 29th, 2018, 3:08 pm
That's interesting about the noise reduction not being useful and that a little room noise is more comfortable to the listener.
-- novel starting here --

I'd like to modify this a little bit. One of my favorite sentences is "Da muss man differenzieren!" - it all depends. On a recording with good equipment like the one you have, I agree, it is not really necessary to do noise reduction. I agree that a little bit of ambience feels more natural, of course, than the generated silence of Audacity (in fact my use of it, if I do, comes out of pure laziness - that way I have defined 5 sec without hassle ;) ). I'd never apply generated silence "inside" a recording, that just sounds ugly. I also agree that it is always better to remove noise on the source if ever possible. But: afair the pros with super-duper-equipment do use equalization, also in use of a kind of noise reduction. The noise reduction of Audacity only is simpler to explain and to use. Pros don't use that program anyway, I guess, but a (real) DAW is more complicated to use and a lot of people here use Audacity, so help is available easily.

When I started to read, good microphones were financially far beyond reach of many, many people. There were tests were we really had to struggle to get the voice out of that noise! The readers read wonderfully, it was a pity, I sometimes was really sad about it and dreaming of hearing that special reader with a better microphone, but they just couldn't afford it.

Times have changed a bit and equipment has become dramatically better at an affordable price. Still there are some people out there who don't have as decent stuff but why should we reject them?

We're here to make available every book in PD as audiobook, no more and no less. There's a quite clear policy about accepting equipment and minimum standard and I am aware that it is not really high. I am happy about every reader with good equipment and good reading. I still love the readers who struggle and put a lot of effort in to learn how to deal with all that stuff out of the pure love of reading and the desire to help and I think they're absolutely worth being supported according to their needs. The ones who can do professional recordings will find their way easily on their own :)

-- novel ending here, thank you for your attention and understanding --

Post Reply