Freesound Forums

Greetings,

There seems to be two broad classes of contributions here, field recordings and samples/mixes. I figure I might as well find out how people tend to use field recordings so I don't unneccesarily limit the usage of my sounds. My perspective is field recording, I haven't got the talent or inclination for sampling. Field recordings, particularly ambiences tend to be longer, and so I mp3 mine, in the assumption that:

field recordings are more likely to be used as sound backings for other productions, probably low in level so mp3 isn't too much of an issue.

sampling seems to be used and fed back more, so mp3 is bad news as artifacts my build up be emphasised, and anyway the clips are short.

But then I don't really know if this assumption is valid. And I've already had one objection to using mp3

http://freesound.iua.upf.edu/samplesViewSingle.php?id=23639

and yet it's so tedious to u/l something 5 times as big. So how much does mp3 impair the usage of field recordings in practice. And, for instance, is not a 10min ambience recording in mp3 easier to use than a 2 minute wav version that you have to loop and risk the loop being obvious?

I obviously record in uncompressed format. flac would be the obvious solution but from other threads seems to limit the number of potential users.

So how big a deal is mp3 compression for users of field recordings? Maybe recordings where there is an obvious subject should be wavs, as these may be used as the basis for a sample, and ambiences are better as longer mp3s? Do most users just hate mp3?

for me, it's obvious, we should not consider any degradation of a sound due to a compression. Because we are talking about hard disk space and bandwith, these two things evolve and become less and less of a problem for the most of us. Maybe freesound could integrate an automatic and compressed version of the sound as a choice for download?

48 khz and 24 bit is my norm for field recording. Because we finish the films at 48-24 at the Final Mix, so we go from beginning to end with these settings. 96 k would be interresting, but really, I don't see the day we will finish a Mix for Films at this rate because
of the processing and hd space requested...

I could see some advantage of using a 96 khz and pitching it down.....down....down and still playback a good sample...

I think the officially preferred method is to upload in a lossless, non-proprietary format, such as FLAC. This saves on bandwidth and disk space (I find that FLAC can usually cut the size of an uncompressed AIFF/WAV in half). Uncompressed AIFF/WAV is probably second preference (although I'm sure it's the first preference for many users), then a lossy format... my guess is high bitrate OGG, then AAC, then MP3, or something like that. So, I would reccomend you try a FLAC converter and see if you can get a decent file size for your field recordings.

I have chosen .flac because no program I use can handle .mp3 without conversion to .wav/.aif anyway, so if I have to convert, it might as well be in a lossless format.

My experience of giving FS users a choice of FLAC or MP3 is that the MP3 gets 2.5 times the number of downloads, but lower scoring (tho' maybe not statistically valid).

e.g. StormLoop22Jul Flac and MP3:

http://freesound.iua.upf.edu/samplesViewSingle.php?id=21043 is MP3, 251 downloads, 12 x 8 stars
http://freesound.iua.upf.edu/samplesViewSingle.php?id=21042 is Flac, 96 downloads, 2 x 9 stars
Both were uploaded at the same time

Similar results with Rain30s.mp3 and .flac, tho' the .flac has been available for longer.

mp3 version 9 x 8 stars
flac version 8 x 10 stars

MP3's at 256 kbps in both cases.

Mike

Hi,

ermine, I fully see your point on sample popularity vs format, but can't agree.

Complexity: the only reason that what we call 'field recordings' tend to be longer is that they include a lot more information and are more complex and unforeseable in their content than man-made samples. Every bit in a sample recorded out there *cannot* be recorded again. Well, this might not hold for repetitive natural sounds (rain falling on a roof, wind noise, ocean waves, etc), but generally speaking the content of a field-recorded sample cannot be foreseen (btw, that's why I love field recordings). In short, every bit and frequency inside a FR sample counts (yeah, I know the high freqs in your starling sample are not perceptible to humans, but for a researcher dealing with bird songs might count, now or in the future)

There is yet another reason that FR should not be downgraded: they can be sampled themselves for a particular fragment that suits the user. I mean, if you upload 3 min of a blackbird singing, you are giving everybody the opportunity to pick the chirp that he/she likes most.

So, the only way to solve the size/quality trade-off is giving the user the choice to download the original OR a compressed version. Unfortunately, and according to Bram, this would impose a severe load on the server. Maybe in the future, with a more powerful machine. Or we could use some form of distributed computing, I don't know.

I don't believe a general strict rule could be established. My opinion is that it's for the uploader to decide the format on the basis of content, complexity, willingness that the sample will become popular, etc

Sorry for the wordy post

saludos

thread related - compression dependent...a certain spectrum should already be available at the fieldrecording...the high and deep tints are decisive...cya

tro

the official position is: we prefer uncompressed, but don't impose anything

this goes for ALL sounds.

for example: what if there's an incredibly good field-recordist out there who's on dialup? should we punish him for not having a fast connection? I think Freesound has always stood for a relatively high signal-to-noise / quality-quantity rating, and I will try to keep it like thiat, but freesound is not supposed to be elitist in any way.

So, I voted "no". It all depends on the situation.

- bram

oh and yes, we could offer high quality mp3 and ogg downloads, but -like dobroide said- this would be tougher on the server, and with over 2Terrabyte/month (almost 10 megabit/second!!) being soaked from the server into the network we would like to focus on other things first ( being, mirrors )

- bram

I voted no. I think the people here who make field recordings are perfectly capable of understanding the benefits of uncompressed sounds. Therefore I think that they have a good reason to release their recordings in MP3 instead of wav or flac. The reason might be bandwith, plug-ins, user-friendliness... whatever the reason, they will have given it a good thought before releasing it as such.

Besides, I stick to two adagia: better a sound than no sound - and one man's thrash is another man's treasure!

LG

> edit: typo

ah, by the way... If you upload mp3's, I think they should be at least 160kbps or better, otherwise the quality REALLY begins to suffer very badly

perhaps it would be cool to do a poll with bitrates: what mp3 bitrate do you consider "enough" for field-recordings: 128, 160, 192, 256 kbps

- bram

Bram
perhaps it would be cool to do a poll with bitrates: what mp3 bitrate do you consider "enough" for field-recordings: 128, 160, 192, 256 kbps

thet were my thoughts also

Bram
128, 160, 192, 256 kbps
- bram

For mp3 I use variable bit rate, with average set to 192. Yields fairly good quality/compression in my opinion.
Oh, of course with Lame Encoder.

I use field recordings just for listening enjoyment, usually while working. I suspect others do this too. e.g. there seems to be a huge interest in thunderstorm recordings. Thus I actually prefer field recordings to be compressed, because I do not edit them.

A while back I uploaded an MP3 field recording. Bram asked if I could also upload the uncompressed version, which I did. Thus far the WAV has been downloaded 82 times and the MP3 91 times, which is nearly the same daily rate given that the WAV was posted a month after the MP3. Thus I voted "no" since people seem to like both (though I agree it is an inefficient use of disk space).

I would guess that it is difficult/dangerous to make assumptions about how people are using the freesounds. Perhaps this survey will help us find out.

daveincamas
Thus far the WAV has been downloaded 82 times and the MP3 91 times, which is nearly the same daily rate given that the WAV was posted a month after the MP3.

Hmmm ... quite strongly weighted in favour of the WAV actually, and different to my experience. But perhaps this is because your MP3 was at 128 kbps, and my MP3 was at 256 kbps. 128 probably too slow and impacting the quality too much, especially for a sample with "white noise" type of sounds which MP3 doesn't like.

SQL: select substring(extension,1,3), avg(nDownloads/datediff(now(), dateAdded)) from audio_file where dateAdded < "2006-06-01" group by substring(extension,1,3);

Result:
+--------------------------+--------------------------------------------+
| substring(extension,1,3) | avg(nDownloads/datediff(now(), dateAdded)) |
+--------------------------+--------------------------------------------+
| aif | 0.389695 |
| fla | 0.319388 |
| mp3 | 0.339362 |
| ogg | 0.340542 |
| wav | 0.381102 |
+--------------------------+--------------------------------------------+

In percentages:
aif 22,01555967
fla 18,0436125
mp3 19,17203033
ogg 19,23869365
wav 21,53010385

Interpretation:
the average nr of downloads per day, for each format.

This means that people really don't care what format the file is in! As you can see the average nr of downloads per day for wave/flac/aif/ogg really aren't any different... I selected only files before 2006-06-01 because otherwise we get skewed results: newer files have higher downloads/day and we are only interested in the stable condition. However results change very little if we remove the date constraint:

aif 19,6578881
fla 19,02992022
mp3 20,18876814
ogg 19,50586116
wav 21,61756238

- bram

Oh and if you feel like giving it a shot yourself:

http://freesound.iua.upf.edu/experimental/downloads_per_extension.txt

collumns (tab-separated) are: ID, extension, nrDownloads, dateAdded
for all files added before 2006-06-01

perhaps someone can make the histograms of nrDownloads/day for each extension.

have fun!!!

- bram

Unless I am very much mistaken, WAV is king...

Plus a general downtrend?

http://upload4.postimage.org/1430857/freesoundls.jpg

you're forgetting two things:

1. there are many more wave files on freesound than flac files. so, obviously there's more downloads for wave files

2. files that have been added later obviously have less downloads!!

- bram

Okay, let's try that again with the brain a little more engaged

I don't think I have a way to account for the greater prevalence of wav files, but scaling for the number of days a file has been up to get the average per day

http://upload4.postimage.org/1431975/freesoundls2.jpg

I have no real explanation for what happened in the last month (May 06). but wav still seems to carry the day. Would have to scale by the percentage of the file formats on the system to correct for that factor, but I don't think I have that info.