Freesound Forums

Hello, I'm new around here

Freesound is an amazing project, which I hope will be sustainable for many years to come. After reading the 2018 sustainability report, I realized that many sounds are in the wav format. Isn't it a bit wasteful on the bandwidth? Wouldn't it be worth it to compress wav files to FLAC (which is lossless, same quality) on the server, and let it be the default download option?

Then, also allow people who can't use the format (or just don't want to) download the original wav file. Then maybe we'd have the best of both worlds.

What does the staff think about this?

Surely your concern about saving server bandwidth is contradicted by your suggestion of having both the FLAC and the original WAV available for download in addition to the transcoded streaming file.

FLAC files are very inconvenient for professionals in audio so it's hard to get people to agree with having FLAC as a standard download here. It's just not supported widely enough for it to be viable. I'm not denying that having lossless audio at a smaller file size isn't superior, i'm saying that right now it's redundant due to software incompatibilities.

I use flac for really large files, but in general if I upload a sound in flac then I get complaints about users not understanding how to use it or telling me to reupload in wav or mp3. The download number of my flac files is considerably lower than wav files.

Agree with all the points made. It is not the first time that a format or technology is not adopted in favour of another that is technically inferior but more popular (for commercial or other reasons). The beta videotape format comes to mind...

Anyway. Be that as it may, for useability Freesound should not make FLAC the default format. It is still supported and people can upload in that format, if they want to.

Personally, I have downloaded some sounds in FLAC format a few times. It is a pain to have to convert these files using additional software, so I can then try the sound in my DAW and see if it fits the project.
But worth it, if the sound is a good one

It's been 4years now, Id expect better software compatibility with FLAC now.
But I'm asking more or less the same thing.

So far I upload in FLAC, because lossless audio at a smaller file size seems just beneficial without any downsides.

Problem is there is a downside, not everyone seems to be able to use it. (Which I'm surprised that not every software supports it these days)
It's relatively easy to convert it back to wav using Audacity or Media Human Audio Converter. But one needs to take the effort of converting and know how to do it in the first place.

So question is to the Freesound staff.
Is that 50-70% file size reduction actually impactful to the sustainability of the service?

Then hard drives are cheaper than ever, at a personal level storing tons of audio files in WAV instead of FLAC is a negligible cost. But I can imagine it being a significant cost at a audio library website scale.

How does Freesound.org position themselves regarding the convenience of WAV vs the storage benefits of FLAC?

Is there a middle ground of "1h recordings should probably be in flac or even lossy .OGG format. But 3s samples are totally fine in wav"? Id then ask where the lines should be, like for 5min recordings? x')

Here I'm not asking Freesound to impose FLAC by converting all audio to it. Just asking what I should do at a personal level and what Freesound.org generally prefers, because at the end of the day it's at their cost.

Cheers!

I try to use 320K MP3 for anything over roughly 2 minutes long. MP3 compression does barely anything to frequencies between 50Hz and 10Khz. Higher frequencies are extremely directional and will always be noisy and out of phase even if you have in-ear monitors. It takes more energy to create the same amplitude at higher frequencies, which introduces noise.

It is easier to lowpass things, record things dark, or high-shelf a dark recording than it is to tame those >10Khz frequencies. Better to psychoacoustically suggest those frequencies than have them be actually present, if you can. If you do this, you will have a very hard time A/B testing flac against mp3 and you will only lose the "ear tickling" frequencies - components of the sound which do nothing for the timbre or harmonic structure and which we lose the ability to hear with age anyway.

Or you can cheat like me and use jb Ferox to constrain the recording to the frequency range of a tape My setting tops out around 14Khz and works for almost everything.

Sadiquecat wrote:
Problem is there is a downside, not everyone seems to be able to use it. [FLAC]

I still occasionally receive complaints about FLAC, e.g.
https://freesound.org/people/Timbre/sounds/708908/#comments

I only ever owned 1 media player that supported flac. It was a Cowon MP3 player from 2005.

I've never seen hardware modular samplers able to load anything but wav.

It's not a pain to convert one file, but if your project needs 12 of them to create a soundscape, or the conversion site puts you in a queue, that could mean half an hour until you get to start carrying out your idea. When people encounter this type of frustration, they will just go back to what works.

Asking people to download codecs or software to support our sounds feels like too much. Many people don't install anything on their PC manually if they can help it. Even if we send them a good converter and it's got 1000 reviews saying it's free and easy to use, they won't do it. The reality is that wav and mp3 are the most accessible and universally compatible formats. If you want things to Just Work for any user and use case, you will use one of them.

It should also be mentioned that Freesound is probably not streaming 24-bit wav to you when you play a 24-bit wav file. They probably use some form of compression to make sounds streamable, like any other site. So, the stream of bandwidth that they send might not vary much between audio formats. This means we really only need concern ourselves with disk space, and only then if we are uploading longform recordings.

strangehorizon wrote:
They probably use some form of compression to make sounds streamable, like any other site. So, the stream of bandwidth that they send might not vary much between audio formats.

That's correct. FYI The preview player is streaming the .mp3 version created in the back-end. Although the bit depth is irrelevant, mp3 doesn't really have a bit depth per se, but the bitrate is around 70kbps.

Doing a bit of digging, I can see that during the processing stage 4 preview files are created; a HQ/LQ .mp3 and a HQ/LQ .ogg. I have no idea why because regardless of the sound the LQ mp3 will be streamed out to the player.

This is for your sound "trash_bag_full_of_cans":
https://cdn.freesound.org/previews/709/709335_1089955-hq.mp3
https://cdn.freesound.org/previews/709/709335_1089955-lq.mp3
https://cdn.freesound.org/previews/709/709335_1089955-hq.ogg
https://cdn.freesound.org/previews/709/709335_1089955-lq.ogg

Sorry if this is too verbose, just thought you'd like to know this info

Please, don't apologize for verbosity. Talking at length about a subject shows respect for it and for those reading it. If we didn't care about a subject, we'd write as little as possible and disregard the thing, right? These posts are up here for posterity so there's every reason to make them comprehensive and educational when we can.

So all those hq files just eat space? Maybe good to batch-delete them.

Anyway I don't mean to derail things away from Sadiquecat's questions.

strangehorizon wrote:
So all those hq files just eat space? Maybe good to batch-delete them.

Anyway I don't mean to derail things away from Sadiquecat's questions.

I think it's right on track for the topic of sustainability

Hi everyone!

As you will know, when your sounds are downloaded their are downloaded in their exact original format/quality (except for the filename, that gets changed). We are not going to change that as this is very much in line with the CC philosophy. Your concerns about end users having trouble with uncommon formats is valid. Your concerns about disk space are also valid, and specially for long files. I think the strategy some of you mentioned to upload short files uncompressed and longer files compressed is a valid one. We don't have any strong preference about where should be the threshold. Also the only impact beyond disk space is the bandwidth needed to download the sound. However, we're not having huge problems with disk space or bandwidth so don't feel too bad about uploading large files

Regarding the format of sound previews, we use low quality mp3 and ogg files (some browsers will prefer one or the other), but also we generate slightly higher quality versions which are not used directly in the website but might be used in some cases through the API (so 3rd party apps that use our API). We are having a discussion with the team about increasing a bit the quality of the previews, but this will definitely have some impact on the bandwidth used. We have to evaluate that, but after all the server changes in the last 2 years which allowed us to have a much faster Freesound with much faster download speeds, I think it will be possible to increase that quality. I'll let you guys know.

As you can see, there are many many things happening around Freesound right know, we'll do our best

I didn't realize FLAC files were so inconvenient for people to work with (I only upload FLAC). I often use an audio converter and find it really easy to use, but I can see how an extra step could be de-motivating for people. Now I'm wondering if I should switch to uploading WAV instead of FLAC (which is sort of ironic, considering the original purpose of this thread).

@frederic.font an increase in sound preview quality would be great!

I recently migrated to a new PC, so the current version of my DAW (Reaper) is able to load FLAC just like it would WAV. I can get the same workflow there. So, when speaking in terms of usability in a professional software environment, any common format is perfectly usable. I have even used AIFF, the bastard son of audio formats, in successful projects - if you can believe that!

However, I insist on using Winamp 2.95 as my media player because it's fast as hell and supports playlists up to 65535 items long. It does support OGG and FLAC (with added plugins), but these formats can crash it, especially if they have improper tags.

All my hardware - field recorders, media players, etc. only support wav or mp3 for recording and playback. So if I want to put someone's discography onto one of them, but it's only in FLAC, I have to mass-convert everything. For me this means making a blank Reaper project, importing the file, then re-rendering it. Perfectly doable, but still less likely to happen. To facilitate creativity, we have to make materials as accessible as possible so there are no stumbling blocks - not even easily addressed ones. Otherwise, we run the risk of destroying the enthusiasm which a user has for a project. Most users do not have a huge variety of solutions in hand like we do, so if they encounter a problem, it could be project-ending. Granted that it's not our fault others are less motivated, but that is the reality and we must give this matter some thought in order to ensure our materials are useful for them.

I do find the tagging system of OGG and FLAC to be inferior to that of MP3 as well. ID3V2 tags can contain any Unicode glyph and work on any software, old and new, without issue. But, I usually only employ tags when the title contains a : or another character which is illegal for filenames, so I've encountered most of my grievances chiefly when listening to others' music.

Finally, this part doesn't matter to me but it matters to many: WAV can embed a number of useful metadata. Loop points, extra info from the author, tempo information, etc. I've never embedded this data into a file myself, but sometimes when importing loops, Reaper can automatically match them to the project tempo, which is quite useful because it gets the tempo right even when the item is a strange length or has a long reverb tail.

@strangehorizon thank you, that gives me a lot to think about!

Honestly, just do what you like to do. Every format has its proponents and haters. And, a lot of what I said about MP3/WAV advantages only applies in specific situations. If you make (for example) drum loops, and they already have tempo information in the filename, and they're already the right length, then all the user has to do is resize them to the bar in their DAW to have them in time. I think Acid WAV and other metadata-bearing forms of WAV were mostly used by early-mid 2000's hardware. I barely see it anymore.