Preserving the USC Optical Sound Effects Library

[Guest blog post by Craig Smith]

Last August, I uploaded 1,233 vintage optical sound effects from various Hollywood productions ranging from the 1930s to 1950s.  This is part of a large project I’ve been working on for nearly three years.  This collection contains many sounds that give us clues about how sound effects were created during the “golden era”.

Here are few examples:

  • Extras on a courtroom set being coached to created dramatic crowd sounds:

  • Western bar fight recorded in sync with camera as production sound, but used as a sound effect many many times since then:

  • A mic in placed in the audience of The Jack Benny Show picks up laughter and applause that can be used for any audience scene:

  • An edited sequence of many sounds from an RKO film that was then used as an effect in low budget films and serials:

  • Leftover wild track of Ginger Rogers laughing for the 1938 RKO musical “Carefree”:

Here is the story of where these sounds came from:

In early 2016, I was working on an experimental western using B-roll found footage from the 1940 film Arizona. This involved my creating a new soundtrack that would sound like it also was from 1940.  My intent was to create a layered track using newer commercial sound effects, then “age” it with digital processing.

But it sounded entirely wrong.  No matter how much noise and distortion I added, it sounded too close and present, and did not “stick” to the image.  It seemed something was wrong with the the sounds I was selecting.  I had to rethink my methods.  After studying the few vintage sound effects I could find, I realized what the problem was.

Until the mid 1950s, there were two ways to record sounds: on a phonograph disk, or on a 35mm optical film track.  Since you can’t edit a record, film sound was nearly always recorded onto film.  The recorders looked like motion picture camera without lenses. The signal from the microphone was used to modulate a light beam which was then was exposed on the edge of the film. The film was then developed at the lab, and copies of it were made for the sound editors. These sounds were hand edited into several “tracks” that would be used for the final mix.

The quality of these original recordings was actually quite good.  The reason these old effects sound different from our new ones is that they were recorded using a different philosophy of where the mic should be, and how much sound should be recorded at once.  For one thing, sound designers were limited in the number of tracks they could create for a mix.  It’s common now to have hundreds of tracks playing back from a computer.  But back then, every track had to be threaded onto a large machine called “dubber”.  Most films probably didn’t use more than a dozen tracks. If you had 12 tracks, you needed 12 dubbers.

Therefore, the recordings were more inclusive.  If we build a western street scene now, we generally add every sonic element individually.  Back then, they would stage the whole scene with people walking, chattering, horse footsteps, wagons, etc.  Then record it with one microphone.  All of it was staged in a way that would support the image, and fit in with whatever sync sound that had been captured during photography.

As for indoor effects, they were usually recorded in large, somewhat reverberant rooms, as opposed to the small, dry, quiet foley stages that are common now.  Once again, perspective was created while recording. The microphone was moved away from the source to create the desired balance between direct and reflected sound.

After realizing that I was using the wrong sounds, I got to work trying to find the right ones.  I contacted several film archives, expecting them to have vast collections of studio sound effects.  To my surprise, they didn’t.  It turns out the studios didn’t really value their sound effects, and the rolls of optical film sound generally were taken by the sound editors as they moved from job to job.  Editors shared and traded sounds with their friends, much like what we do with freesound.org.

The last place I tried was the USC Hugh M. Hefner Moving Image Archive. I went to the USC film school in the 1970s, and remembered there were a lot of old noisy effects in the Sound Department.  Archivist Dino Everett told me that yes, they did still have those effects, but they were about to throw them out.  Did I want them?  Yes!

Not knowing what I was getting into, I went to USC and picked up 12 bankers boxes of 1/4” tapes.  These tapes were first generation full-track recordings made in the early 1970s. USC Sound TAs would thread up small bits of 35mm optical film, then transfer each one to tape. Most of the tapes are in good condition, but unfortunately, most of the metadata (what the effects were and where they came from) is missing.  While asking around to see if anyone might have a copy of the metadata, I found out that the Gold Library portion of the collection (the effects I have uploaded to freesound.org) had originally been transferred by Ben Burtt.  Ben did this as a summer job while he was a USC Sound TA.  Soon after graduating, Ben was hired by George Lukas to design sounds for the first Star Wars film.

I contacted Ben and discovered that he still had his hand-written notes from when he did all those transfers!  (Moral: Never throw anything away.)  Ben generously made me a copy of these notes, and I was ready to start.

Except, of course, I had no idea how much time it would take to transfer, edit, and label all these effects.  I do have a full-time job, and a family, so it hasn’t been easy.  The tapes were transferred from a Nagra IV-L recorder into Pro Tools at 96 kHz, 32-bit floating point.  They were then separated into individual files, and individual effects were level adjusted to -24 dB LKFS.

Typing in all that metadata was the biggest hurdle.  After much effort, I got a couple of grants that allowed me to hire an assistant and buy supplies.  And Soundly gave a very generous in-kind grant of cloud space, and their software, which was used to create and manage the metadata. (Checkout Soundly’s software if you haven’t. Besides being a great way to organize your sounds, it connects to and searches Freesound.

In August, after all the help I received, I was able to upload the 1,233 sounds of the Gold Library to Freesound (tip: I used the new bulk description feature which really saved me a lot of time). Since then, I have been receiving comments almost every day.  This makes me so happy!  The next installment, the Red Library, will hopefully go up next summer. My aim in all this is to make sure these sound effects don’t disappear again.  So keep using them, and keep sharing them.

I couldn’t have done this my self. I want to thank Dino Everett, Ben Burtt, Leanna Kaiser, Randy Haberkamp, Andrew Kim, Peder Jørgensen, Christian Schaanning, Frederic Font, and Lynn Becker.

—Craig Smith

 


Craig Smith has been recording and manipulating sound since 1964. After graduating from USC’s School of Cinematic Arts, he worked as a sound editor and production mixer in Hollywood, specializing in noisy action-adventure films that are blamed for the downfall of society. He left that world in 1986 to teach sound in the School of Film/Video at California Institute of the Arts, where he is now Academic Sound Coordinator.

Craig’s own work experiments with implied narrative and accidental sound design – putting together sounds & images that have nothing to do with each other to create unexpected stories.

Craig is a member of the Society of Motion Picture and Television Engineers, and the Audio Engineering Society.

 

Posted in Uncategorized | 1 Comment

Community update September 2018

Hi again…

…and welcome back to our community update post! This month we don’t have any major user-ready features to show, but there are a number of things we’ve been doing in the background which will allow nice new stuff in the future. Here is a list of things:

  • New analysis of sounds with Audio Commons audio analysis tools. For the last 3 years we’ve been coordinating a project called Audio Commons (see original post in this blog here), focused on developing audio analysis tools and promoting CC audio. We have now analysed Freesound with some tools developed in the project. The analysis results are already available and usable through the Freesound API, but still are experimental and work in progress. This analysis will allow us in a near future to deploy new features, specially in the search page. For example, would you like to be able to filter search results by things like “only sounds including 1 sound event”, or “sounds isolated from background noise” or “only bright sounds”? Well, we have many ideas in mind but these are some of the features we’re planning to research and deploy to Freesound in the near future.
  • Improve tracking of email addresses that bounce. This is important because it will allow us to stop sending emails to addresses that no longer exist and reduce problems with blocked emails.
  • Other minor bug fixes, security upgrades and optimizations.

Also we’d like to take this opportunity to give a quick update about the new interface design that we announced almost 1 year ago. We never gave an estimated release date because we knew these things sometimes tend to require more time than planned, and we did good 🙂 It has happened that even though the design phase has been finished for a number of months now (and we’re really happy with it!), the implementation has been very slow and we’re are not yet ready to give any estimate other than “at some point during 2019”. But yeah, it is coming! Aaaaand as a prize for your patient waiting here you have a little teaser so you can have an idea of the main design lines:

That’s it for now, enjoy Freesound!

The Freesound Team

Posted in Uncategorized | 4 Comments

Community update August 2018

Hi everyone,

Summer’s been hot in Barcelona but we’ve been working hard (well, we did some holidays as well!) and we have some new things deployed to Freesound 🙂 These are the new main things:

  • “Streets” layer in maps: some of you reported that the new maps we introduced a couple of months ago missed the streets layer with street names, etc., and told us that this made your geotagging workflow much harder. Well, we’ve added now a streets layer that we hope will be useful to all of you 🙂

  • Fav icon: believe it or not, we did not have a proper fav icon until now… Thanks superanton for your contribution!

As usual, we have also been working and released a number of other fixes and improvements:

  • Improved spam reports which will help dealing with spam faster (for admins)
  • Small improvements in moderation interface
  • Better monitoring of  our email sending
  • Improvements in invalidation of HTML caches
  • Other minor bug fixes and optimizations

Do you remember a couple of months ago we released the Bulk upload feature? Well, we’re starting to harvest its fruits as you can see by checking this amazing collection of ~1200 sounds uploaded by craigsmith 🙂 Thanks for these great contributions Craig!

That’s it for now, enjoy Freesound!

 

The Freesound Team

Posted in Uncategorized | 1 Comment

Community update June 2018: Bulk description

Dear Freesounders,

Welcome back to the Freesound-dev world monthly post! This month we are very excited to let you know that we finally released a long awaited feature that we think hardcore sound uploaders will love. And possibly non-hardcore uploaders will love as well… hmm and downloaders too!

  • Bulk sound description: we have just added an option in the describe page to let you describe many sounds at once by uploading a data file in CSV or Excel  formats. In this way, you can prepare the descriptions of all your sounds offline in your  computer and using any spreadsheet editing software you like. Then, when you’re ready, save descriptions to a file, upload it to Freesound, and sit down to watch how your sounds are described automatically and appear on the site 🙂  You’ll still need to first upload your audio files to Freesound normally, but this is really easy as the upload page allows you to drag all the sounds you want and upload them at once. Checkout this FAQ entry to get a better idea of how the process works.  I just used it now to upload this pack of 508 sounds, and it worked great!

By adding this feature we expect to make life easier for Freesound uploaders (and others who are not yet uploaders) that have collections of sounds that have not been uploaded to Freesound because it would take too much time. Now you have no excuse 😉

Note that bulk sound description is currently only available for uploaders that have contributed 40 sounds or more. Therefore, if you want to use that feature and are still below the threshold, you’ll have to train yourselves a bit more on uploading sounds 🙂

Even though we tested the bulk description feature quite a lot, it is a big change and it could happen that some things do not work as expected. Just in case, we opened this forum thread so you can leave comments and let us know how it works.

We’ve been also been busy as usual with lots of internal improvements in our code which prepare Freesound for the future. For those curious, remember that you can follow the development at our public code repository.

Aaaaaaan that’s it for now, time to get your data files ready and upload all those great sound collections that have been sitting for ages in lonely hard drives 🙂

Enjoy Freesound!

The Freesound Team

Posted in Uncategorized | 1 Comment

Community update May 2018: Maps

Dear Freesounders,

Here we are back with some more news about things happening in the Freesound dev world! We have just released a couple of new features and have been working on a big third new feature but that won’t finally make it until next month. These are the  newly introduced features:

  • New maps: we’ve just got brand new fresh maps! We switched to a new maps provider called Mapbox (which uses some imagery data from OpenStreetMap) and have taken the opportunity to add some improvements to the way maps work in Freesound. If you go to the sounds map page you’ll notice that, besides the new icons for sounds and clusters, the maps now show terrain labels to easily locate places, and there is a search bar that you can use to enter locations and fly to these in the map. That’s a huge usability improvement as in old maps you had to manually navigate to whatever place in the world you wanted to hear sounds from 🙂 This search bar has also been added to the small map that you use to add geotags when describing your sounds.Together with these improvements we also improved map share links and map embeds. To share map with a link, you can now simply copy the link from the address bar while you’re browsing in the map page. Map embeds now have extra options that allow you to choose the size of the embed and to decide whether you want sounds to be clustered (or not). This is particularly useful for projects that use Freesound map embeds like Sons de Barcelona.

Sounds of Amsterdam

Sounds of New Zealand

  • Simpler registration page: believe it or not, we spend a lot of time answering support requests from users that have trouble registering. We detected that our login form was a bit problematic sometimes gave rather confusing messages about the status of the registration process. That’s why we decided to fix it and we have now a brand new and much simpler form that hopefully will reduce the number of support requests we get from registration. Ah, and we also took the opportunity to get rid of the first name and last name fields which we actually never used and don’t store anymore 🙂

As usual, we have also been working and released a number of fixes and improvements:

  • Fixes in moderation pages
  • Improve cache invalidation for sound pages: to save some efforts to our servers we use caching. This means that there are portions of the HTML we render that are not computed every time you request them but that are already cached and reused for everyone. It can happen that sometimes the cached pages go out of sync with the real data in Freesound. We have improved the mechanisms to invalidate cache so that these will be better in sync with Freesound.
  • Fix a bug in sound comment email notifications which in some situations was ignoring user email preferences.
  • Improved sync between our stored counts of downloads per user and the actual real numbers.

That’s it for now, enjoy Freesound!

 

The Freesound Team

Posted in Uncategorized | 3 Comments

Introducing Freesound Datasets (and more!)

Dear Freesounders,

Today we are very happy to introduce you to Freesound Datasets, a new platform that we’ve been developing during the last year to foster the re-use of Freesound content in research contexts and that will eventually help us make Freesound better and better. Curious? Check out the website at https://datasets.freesound.org/.

But what exactly is a dataset? To say it short, a dataset is a collection of items (sounds) annotated with labels chosen from a limited vocabulary of concepts. Well-curated datasets are one of the most important things that are needed to advance research in many fields, including sound and music related research.

Freesound Datasets is a platform that allows users to explore the contents of datasets made with Freesound sounds. But even more importantly, Freesound Datasets allows anyone to help make the datasets better by providing new annotations. Furthermore, it also promotes discussion about the datasets that it hosts, and allows (or better said, will allow) anyone to download different timestamped versions of them. If you’d like a more academic description about the platform, you can check out this paper we presented at the International Society for Music Information Retrieval Conference last year: Freesound Datasets: A Platform for the Creation of Open Audio Datasets.

Using Freesound Datasets, we already started creating a first dataset which we called FSD. FSD is a big, general-purpose dataset composed of Freesound content and annotated with labels from Google’s AudioSet Ontology (a vocabulary of more than 600 sound classes). Currently, FSD is still much smaller than what we would like, but we are sure with the help of people all around the world it will get bigger and bigger. Needless to say, you are more than welcome to contribute to it (or in other words, please contribute!). All you need to do is visit the Freesound Datasets website and click on Get started with our annotation tasks! We will simply ask you to listen to some sounds and have fun 🙂 You’ll see an interface like this (you can login with your Freesound credentials):

That’s really cool, isn’t it!?

Yeah that’s awesome, take me to this interface because I can’t wait any longer to start annotating!

But you know what? There is even more! We have been awarded a Google Faculty Research Award to support the development of Freesound Datasets and FSD, and, in relation to that, have started a collaboration with some colleagues from Google’s Machine Perception Team to do research on machine listening. As the first outcome of this collaboration, we recently launched a competition in Kaggle  (see Freesound General-Purpose Audio Tagging Challenge), in which participants are challenged to build artificial intelligence algorithms able to recognize 41 diverse categories of everyday sounds. The dataset used for this competition is a small subset of FSD.

The great great great thing is that the outcomes of all these research efforts will help us improve Freesound in many ways. By training our search engine with FSD, we would, for example, be able to find search results inside sounds (for example, a fragment of a field recording with bird chirps), or be able to allow you to browse Freesound sounds using a hierarchical structure. This, and many other things that we will find out in the future 🙂

That’s it for now, thanks for reading…

the Freesound Team

Posted in Uncategorized | 8 Comments

Community update March 2018

Dear Freesounders,

Here is our monthly list of things happening in the Freesound dev world. Again, most of the deployed things are rather small improvements:

  • Updates in the moderation interface and in the “sounds pending moderation” page.
  • Disallow use of symbol @ in new usernames.
  • In the browse by geotags page tag filters are now case-insensitive (e.g. https://freesound.org/browse/geotags/Ship/ shows the same as https://freesound.org/browse/geotags/ship/).
  • Treat email addresses as case-insensitive.
  • Improvements in the API documentation.
  • Internal improvements (quite important!) in the way we store downloads information.
  • Other minor bugfixes and optimizations.

Let’s see if by next month we can release one of the big new features that we are developing in parallel to all of the smaller changes and the development of the new frontend 😉 This one coming will hopefully make uploaders very happy! Stay tuned for the next post…

The Freesound Team

Posted in Uncategorized | 3 Comments

Community update February 2018

Dear all,

Welcome to our monthly update about things happening in the Freesound dev world! We have no significant new user-facing features to showcase this month but, as usual, we’ve been working in the background and been making some backend changes and developing some stuff which is not yet ready to be launched (but will be shortly!). We also continue to advance (maybe a bit slowly than expected) in the implementation of the new font-end interface 🙂 Here is the list of updates/fixes for this month:

  • Store the sounds that are downloaded every time a pack is downloaded. This will allow us to tell you which sounds (and with which licenses) were inside a pack when you downloaded it.
  • Change the way we store the count of downloads that you make.
  • New anti-spam measures to prevent spammers from using “about” and “URL” profile fields for spamming.
  • Currently we show a warning when including images from non-HTTPS locations in sound descriptions. Now this warning is extended to other fields that also support including images.
  • Improvements in the API documentation.
  • Other minor bugfixes and optimizations.

Aaaaand that’s for February. Thanks for reading and stay tuned for the next post!

The Freesound Team

Posted in Uncategorized | 2 Comments

Community update January 2018

Dear Freesounders,

Welcome back to our monthly community update post in which we tell you about the things that have been happening in the Freesound dev world. This month we are releasing a major new feature and, as usual, many smaller things which improve Freesound here and there and make necessary steps for future improvements and new features. Our big new feature this month is…

  • Username change: from time to time we receive requests from users that want to change their usernames. Until now this was not available as an option and we only did it manually after support requests. From now on, you’ll be able to change your username from your settings page. Even if you change your username, posted links to your sounds, packs, etc., will still work (we redirect to the new username). Nevertheless, please don’t go and start changing all your usernames now, 99% of your usernames are certainly really good!  This is only intended for special occasions. Also, note that you can only change your username a number of times and that, when changed, you won’t be able to go back to your previous one.

Besides this new features, here is a list of other updates and fixes:

  • Store all the times you download the same sound as different records. In this way, in the future we’ll be able to tell you in which dates you downloaded a specific sound and with which license.
  • Improvements in our admin interface to manage API applications.
  • Updates in the documentation of the API.
  • Improved processing of HTML content in forum posts, sound/pack descriptions, about field and forum/sound signature fields.
  • Add a link to the sound moderation ticket from the sound page (only if you are the author of the sound or a moderator).
  • Other minor bugfixes and optimizations.

Aaaaand that’s for January. Thanks for reading and stay tuned for the next post! Some new big things are coming for uploaders 🙂

The Freesound Team

Posted in Uncategorized | 5 Comments

2017 in numbers

Dear all,

Just like we did one year ago, we have collected some statistics of Freesound usage during 2017 and we’d like to share them with you. The stats will be presented following the same structure as in last year’s post so let’s get started with the number of new sounds uploaded during 2017:

36,517 new sounds!

which corresponds to…

661 hours of audio!

This means that this year there have been around 1,000 more new sounds uploaded than in 2016. Interestingly, sounds uploaded in 2017 are 5 seconds longer in average than sounds uploaded in 2016, thus the increase in hours of audio with respect to 2016 (73 hours) is bigger than expected.

This year the vast majority of uploaded sounds were released under the Creative Commons 0 license, as opposed to last year where the Creative Commons 0 and Creative Commons Attribution licenses were equally distributed:

Freesound currently hosts more than 363k sounds, which means that it would take you a full 213 (and a half) days to listen to all of Freesound 😉 We still have not reached the mark of 400k sounds, but if you continue uploading sounds at the same rate it’s just a matter of months until we get there…

If we look at the tagcloud (see below) of the uploaded sounds we’ll see that, again, field-recording is the most used tag of 2017. However we see also other tags which are very common: vsco-2, multisample and single-note among others. This is because of the addition of the Versilian Studios Chamber Orchestra 2 Community Edition library of ~3,000 instrument samples that happened back in January 2017.

And now getting to the ranking of uploaders. Last year there was a bit of controversy because we showed a raking of users sorted by the number of uploaded sounds during 2016, but this did not take into account the duration of the sounds. So this year we are showing two rankings, one ordered by number of uploads and the other ordered by total uploaded time:

Username # uploaded sounds Username uploaded time (hours)
#1 Samulis 3,071 #1 klankbeeld 54
#2 kyles 655 #2 shuraifa 47
#3 penetrermind 583 #3 kyles 27
#4 GCGuest1 536 #4 csengeri 27
#5 reklamacja 458 #5 awaka 23
#6 pjcohen 436 #6 felix.blume 14
#7 Jovica 366 #7 maqsood5@mail2world.com 8
#8 deleted_user_2195044 351 #8 JonnyThePonny 7
#9 ldezem 321 #9 InspectorJ 6
#10 Glitchedtones 319 #10 DunyaninSesleri 6
#11 InspectorJ 311 #11 Corsica_S 5
#12 klankbeeld 291 #12 frambo44 5
#13 eardeer 291 #13 ChrisReierson 5
#14 Anthousai 285 #14 ColdWombDescent 4
#15 felix.blume 268 #15 BeeProductive 4
#16 deleted_user_7146007 263 #16 deleted_user_7146007 4
#17 cabled_mess 253 #17 Jovica 4
#18 awaka 242 #18 gis_sweden 4
#19 lomographicmusic 233 #19 Samulis 4
#20 jalastram 230 #20 dbspin 4

As you can see, the number of uploaded sounds and the total number of uploaded hours are different beasts, but there are some users who made it into both rankings. Congratulations to all uploaders (and moderators!) regardless of whether you are or not in the ranking!

Let’s now continue with statistics about the number of downloads. The number of sound downloads (including packs) during 2017 was…

16,600,000 downloads!

So yes, you’ve beaten last year’s record of downloads. In fact, Freesounders have made more than 110M downloads in all its history, which means an average of 300 downloads per sound. Quite a number! Nevertheless, if we look closely at the downloads data, we see that there are a few sounds which are downloaded A LOT, and MANY sounds which are not downloaded much. In statistics, this is called a long tail distribution, and it’s something very common. Here we show a histogram of the number of downloads per sound:

This histogram should be read like “40% of the sounds in Freesound have been downloaded between 0 and 20 times”, “10% have been downloaded between 20 and 40 times”, and so on. In this histogram we only show up to 500 downloads per sound, but to make it complete we should show until ~130,000 downloads per sound (which is the number of downloads of the most downloaded sound in Freesound). The global average is at 300 downloads/sound, but the the likelihood of getting 300 or more downloads after uploading a sound in Freesound is not as high as you would expect (only one out of six sounds get, at least, that many downloads!). However don’t worry too much, 96% of sounds get at least one download 😉 In any case, you know what you have to do: upload awesome sounds! But before downloads happen, people need to search for stuff, so let’s see what Freesound users are searching for:

Does this look familiar? It looks quite similar to the termcloud we showed one year ago doesn’t it? Well, apparently wind, explosion, music and whoosh (in all it’s spelling variants) are still the most commonly used search terms. For example, on average for every day, there are more than 400 queries with the term wind and more than 270 queries with the term explosion. Other terms in the termcloud like scream or footsteps also seem to be pretty stable from one year to the next. We’ll see what happens in 2018!

To finalize this post, some numbers about messages, ratings, forums and comments. During 2017, you’ve sent more than 19k messages, added 1.4M sound ratings, written 1,900 forum posts and made more than 43k sound comments! With the exception of sound ratings (which has increased), these numbers are slightly lower than last year’s stats. We’ll see what happens next year after the new interface has been released and the impact that it will have on all these numbers.

Thanks for reading and enjoy a 2018 full of sounds!

The Freesound Team

Posted in Uncategorized | Comments Off on 2017 in numbers