Transcript TITLE
Music streaming and downloading:
an introduction
Richard Ranft
The British Library Sound Archive
IAML Annual Study Weekend – April 2006
Principles: file sizes,
audiences, copyrights
Audio file types
Preparing sound files
Delivery methods
Examples
2
2
Digital audio
“Music is a time-frequency wave function with tonal
qualities pleasing to the human ear” – ALICE
Digital audio files are large…
In a simple text file (Windows Notepad), one character or
space requires 1 byte storage
The 325 letters and spaces on this page occupies 325
bytes
This same text as recorded speech = 1,000 x larger
3
3
File sizes & data rates
8 bits = I byte
Data storage sizes measured in bytes
CD 650 MB
1 minute @ CD quality ~ 10MBytes
Typical desktop computer hard drive 60 GBytes
Data transfer rates measured in bits per second
CD quality stereo
1,411 kbit/s
Modem connection
up to 56 kbit/s
Broadband connection
from 256 kbit/s
LAN
from 10 Mbit/s
4
4
A quart into a pint pot
Downloading
Playback cannot start until entire file received by user
Playback is continuous (no buffering)
Transfer high quality, uncompressed files
Compression
Lossless compression, e.g. MLP, FLAC, 50% reduction
Lossy compression (“data reduction”), e.g. mp3, AAC, RealAudio, >80%
reduction
Used for streaming media
Streaming
Fast delivery, but lower quality than downloading
Useful for previewing
Instant playback: plays when you click
Can use to discourage copying
Buffering: used to control network interruptions
Live streaming used in ‘broadcasting’ radio shows
5
5
Your audience
Why deliver sound?
Who are your
audiences?
What content do want to
deliver?
How to deliver content?
Accessing the content
Rights issues
6
6
File types
Uncompressed files
WAV
AIFF
3-minute song (32 MB): several hours to download over modem
Compressed files
MPEG-1 layer III .mp3 (typically 10:1 compression at 128 kbps)
Ogg Vorbis .ogg
Quicktime
Windows Media Audio .wma
Choice of bitrates
7
7
Preparing files
Master files v. access files
Digitisation & editing
Use high quality files during preparation
Edit with e.g. Wavelab (Steinberg), Adobe Audition (Adobe),
SoundForge (Sony), Audacity (freeware)
Conversion from WAV to access version file
Batch conversion tools
Batch renaming e.g. Rename4u
Upload to server
8
8
Delivery
Simple web page
Choice of file types and qualities
Streaming & streaming servers
Podcasts
Preventing piracy:
Use short samples or low resolution
Watermarking, encryption, streaming
9
9
www.arkive.org
3 file types offerred
(with video), each
at 2 bitrates
10
10
Click to add title
Cylinder Preservation
& Digitization Project:
http://cylinders.library.
ucsb.edu
3 file types offerred:
Quicktime for
streaming; mp3 and
wav for downloads
11
11
Examples
British Library Sound Archive
5 million recordings: musical works,
interviews, recordings of plays and
poetry, wildlife sounds, oral history,
sound effects
580,000 hours (excluding second
copies)
12
12
66 years non-stop listening
6 years added per year
Most of the recordings are
in copyright
13
13
Access
Access: CD publications,
exhibitions & talks, radio
broadcasts, copying service;
listening services
- where rights can be cleared,
or where BL owns rights
14
14
Listening access in three arenas:
Off site, all web users, via BL
website
Off site, selected web users, to
FE & HE communities in the UK
Collect Britain (1,400 sounds), plus 800
other sounds, where copyright
permissions have been granted or are in
public domain (older than 70 years).
Archival Sound Recordings Project:
12,000 sounds by September 2006.
On-site access
unrestricted (3 million sounds…)
15
15
www access to British Library sounds
Before 2006: RealAudio streaming format
Now: WMA streaming using MMS
RealServer
Surestream
Windows 2000 server software
45 simultaneous users, but can accommodate many more
400 GB server storage capacity
Downloads not yet available
Mp3 downloads of Archival Sound Recordings Project
recordings from September 2006
16
16
www.bl.uk/nsa
17
17
18
18
On-site access
For IPR reasons, continuing requirement
to provide on-site access
Manual Listening & Viewing Service,
appointments-based
Many recordings too fragile to be
handled by listeners
Headphone listening in soundproof study
carrels
Limited number of carrels
Cannot meet increasing demand from
playwrights, media researchers, oral
historians, musicologists, etc
Risk of handling damage
19
19
SoundServer
A new service, providing instant
access to many sounds in selected
Reading Rooms
Access to digital surrogates of
recordings stored on webserver
‘Point & Click’ delivery from selfservice workstations in reading rooms
Use of ‘leak-proof’ self-muting
headphones on selected Reading
Room terminals
20
20
How does SoundServer work?
Staff receive request for CD listening
On-demand encoding to mp3 files
Files transferred to Windows 2003 server
Personal playlist created by staff using a web form
Playlist links to mp3s and to Sound Archive catalogue
Listeners wait 5 working days for pre-ordered custom playlists
Listener accesses their personal list via Reader number
Sounds relayed over 100 Mbps network to Media Player Classic
player on web browser
21
21
SoundServer: benefits
Removes need for appointments
Semi-automated service - listener-controlled playback
Potentially many simultaneous users
Increasing on-line data bank of audio
Reduced wear on collection items
Entire SoundServer collection accessible to any visitor via
‘Jukebox’ pages
Runs in parallel with existing appointments service, using
existing staff resources
Low cost: capital costs of £14k for server store, headphones &
2 workstations, but rest of service uses existing simple, lowcost technology, & existing staff
22
22
SoundServer: disadvantages
Lower sound quality of surrogates – but acceptable to
most users
Highly dependent on IT infrastructure/reliable network
to add and access content
Users wishing to read sleeve notes and view artwork
must still make an appointment – but audio-only
access acceptable
Need to continue existing appointments-based manual
service in parallel for ‘difficult’ audio formats and for
sleeve notes and video viewing.
23
23
Progress so far
Nine PC terminals
Currently limited to CD
recordings
A growing collection - now
3,000 recordings accessible by any readers
browsing the Jukebox
- including our most
frequently requested items
Feedback from listeners is
very positive
24
24
SoundServer developments
Extra
SoundServer terminals in reading rooms and in
public areas of building
100,000
Adding
files in 4 years
e.g. BL CD publications: ‘try before you buy’
Integration
with other electronic audio resources, other
BL electronic resources (e.g. Oral History transcripts,
Naxos on-line library)
Introduce
improved playback facilities (bookmarking,
saved searches)
Submitting
playlist requests via a webform
25
25
26
26
27
27
28
28
Growth in listening and viewing use
140
number/week
120
L&V appointmentsexcluding SoundServer
100
80
60
Weekly L&V (inc. SS)
totals
40
20
0
07/08/05
07/10/05
07/12/05
07/02/06
07/04/06
date
29
29
Click to add title
Thanks for listening.
Questions?
30
30