We've sent a verification link by email
Didn't receive the email? Check your Spam folder, it may have been caught by a filter. If you still don't see it, you can resend the verification email.
Started May 21st, 2008 · 36 replies · Latest reply by qubodup 14 years, 7 months ago
Hello,
for previews and waveform/spectral displays only, I was thinking of doing peak-to-0dbfs normalization on all files.
good idea?
advantage: it shows the waveform in moredetail for very silent sounds
disadvantage: the previews could become more noisy for silent sounds
- bram
the only problem that may result with a volume control is it may only be able to decrease the volume of a sound , not boost it any. Most flash based players I have used play the sound at unity when it is at maximum volume and attenuate from there. If it can be done (BRAM the MAN) it would be sweet.
Normalizing would be cool. Yeah, it would show the defects of very low level recordings before you went and downloaded them. I think this would save some people a lot of hassle once they download the sound and realize it's not very useful to them at the low level it was recorded at.
remember I'm talking about the volume of the preview and normalization of the display:
( the same file, the second one is normalized in the display )
http://freesound.iua.upf.edu/data/38/images/38660__ERH__bell_sequence_5b.png
versus
http://freesound.iua.upf.edu/files/wav2png/38660_ERH_bell_sequence_5b_w.png
+ of course the volume change in the preview.
- bram
Yes,
BUT the first things I see is that the colors are changing,
so what you can do is to do the regular processing
but before drawing the peaks
you just normalize that peak array
and AFTER you draw them
so the result is that the colors
give the real spectral information
in the native state of the sound
and itS more easy to see small waveform
and the processing take less time,
cuz there is less peaks value than sound samples value to process
I would choose not normalized because it is easier for me to tell what the resulting sound will be like in terms of volume and dynamics. Your normalized example suddenly looks very loud to me, even though it is not, and the resulting increase in frequency resolution, doesn't really tell me much more about the sound than the non normalized example.
-Richard
RHumphries
I would choose not normalized because it is easier for me to tell what the resulting sound will be like in terms of volume and dynamics. Your normalized example suddenly looks very loud to me, even though it is not, and the resulting increase in frequency resolution, doesn't really tell me much more about the sound than the non normalized example.-Richard
The reason I found this site is, I was looking for a sample of wind(storm), and so I googled 'wind.wav'...anyway, once I got started looking through the site, it was important to me to be able to look at the fantastic graphical images of the wave that the site incorperates, at a glance, I could see where the wind was soft and quiet, and also where it built up to gale force. I then started looking for thunder sounds, and the same graphical representation allowed me to be able to find samples of light(ambient) thunder...or heavy close range crashes.
When you have an image of a required sound in your head, you then tend to have an image of the sort of wave you're looking for, and this is where the 'un-normalized' images are most useful.
It is a fantastic idea to normalise the waves but for me, I think it is perfect the way it is...just like RHumphries said above, I tend to 'read' the wave before I listen to the sound, and surely normalising the samples, would actually make life more difficult.
I would just like to add(as this is my first real post), that this site is absolutely fantastic, and has opened up new horizons for me.
Keep up the excellent work lads. :wink:
I voted normalize. I really use my ear more than my eyes. Chalk it up to inexperience I guess.
My thought would be the normalization would give me a better way of quickly judging a sounds potential on something quiet.
EJ fortin's idea would satisfy both angles of keeping the original wav image but normalizing the preview for our ears to judge...
The changed colors are just because I changed the palette of the colors used for the sounds. I thought they looked so bad I had to change them for the nightingale!!
Check (sinusoid rising sweep) new:
http://freesound.iua.upf.edu/files/wav2png/11_Bram_sweep_log_w.png
versus Old:
http://freesound.iua.upf.edu/data/0/images/11__Bram__sweep_log.png
The same information is there, it's just different (nicer!!!) colors.
For the spectrogram I need to normalize the sound, but as in the spectrograms the amplitude of the sound is not so visible it doesn't matter.
The spectrogram will will be displayed on mouse hover in the sound detail page.
I guess Richard is right though: the normalized wave-display doesn't tell us more about the sound, it tells us less because you can't guess if the original sound is silent or not.
But ejfortin is correct when he says flash can only DECREASE volume, not increase.
Richard, what do you think about normalizing the mp3 preview?
I'm still doubting here!!
I should really make an example page where I combine
http://freesound.iua.upf.edu/files/player/player.html
with
http://freesound.iua.upf.edu/blog/?p=10
so you guys can see the final result.
Mart1001, all (well explained) opinions count!! That includes yours
- bram
I think ur right.
Normalising the audio gives you less information. The graph should be a visual representation of the audio you are downloading.
If the visual (and audio) is normalised I think this is confusing and not helpful.
Also personally I don't prefer either set of colours Both look fine to me.
Bram
Mart1001, all (well explained) opinions count!! That includes yours
I know it's only the preview we're talking about but, that's all I tend to listen to initially. I shouldn't have to d-load a sound before I can rate it. For example, 'the random sound of the day' on the home page, 'rate this sound', I should be able to just preview it, and then rate it, knowing that the sample I'm rating, is more or less what I'm hearing without any workshop aid such as normalise.
Surely, it's upto the technician to ensure that he gets his sample spot on before submitting, in order to obtain the best possible rating. Of course...some sounds need to be quiet else they lose that feeling of realism...and normalizing may also make a classy sample...sound amateur.
I wouldn't want my recording of a snail, to sound like someone dragging a wooden box across the floor.
Sorry...I'll shut up now.
Bram
<snip>
Richard, what do you think about normalizing the mp3 preview?
<snip>
- bram
I can see how that would be helpful, but my gut instinct would be to leave it alone since normalizing the preview would not give an accurate impression of the sound. But this is all really just my personal preference-- I'm not speaking with any sort of authority. If everyone else would prefer it normalized I could live with that too.
-Richard
After some thinking I voted no to normalizing,. Look at it this way: imagine you are exploring a vast library filled with bookshelves in which ALL books are exactly the same size (someone took the trouble to normalize the books). Would this make easier or harder to find a particular item? It might be argued that size is a superficial characteristic , it doesn't matter because books have spines with colors and letters (their 'spectrograms') that identify them, and it's true. But you need to look closely to read a book spine...
You may say 'I'm the tidy kind of a librarian, books look nicer the same size'. Allright but... a girl once told me size matters. Small sounds... sorry, I mean quiet sounds can be easily enlarged... ooops amplified to a comfortable loudness during preview play.
You know, this normalization issue makes me think of an email I got the other day, can body parts be artiffcially increased in size? Really?
D
firefreak, mouse-over will be used to display the spectrogram of the sound ( see http://freesound.iua.upf.edu/blog/ )
I think dobroide, Richard and Mart1001 have convinced me.
By the way, the sidebar on the site will disappear (too much text scares the new users!) and all waveform/spectral displays will be a bit larger (like the ones you are seeing here).
But that is a matter for another thread.
- bram
Bramadvantage: it shows the waveform in moredetail for very silent sounds
disadvantage: the previews could become more noisy for silent sounds
not normalized, but the nightingale player with the new previews and spectral display on mouse over:
http://freesound.iua.upf.edu/files/player/player.html
opinions?
I think the spectral display should probably get a button ( small S perhaps? ) instead of on mouse over.
- bram