 | Craig Meyer
to Adam
More options Aug 28 (2 days ago)
Adam,
You've been doing a great job promoting Indie podcasting. Thanks
for all your efforts. I always enjoy the "Daily SourceCode", it
makes my commute most enjoyable. I usually arrive at work (or home,
depending on when you post) with a smile on my face, and/or quietly
giggling to myself.
I've been kicking around an idea for reducing bandwidth usage
for certain style podcasts. In your recent (#228, 229) podcasts you
were talking about MushUps, and something clicked in my head.
The quick overview: Generate (assemble) the MashUp on the Users
machine. Rather than sending the final MP3 product, send the
instructions to a MP3 ReGenerator to build the MashUp from the
constituent sound files using a file (stream) of MP3 Assembly
instructions.
So the User would use iTunes (AllOfMP3.com ;) or existing copies
of the MP3 songs, with the MashUp instruction file. A new program
would then stitch together the MashUp on their machine and either
output to the speakers directly or create an MP3 file for the User to
listen to. Assuming the final User has legal copies of the Original
music, the MashUp created on their machine is all kosher! This would
generate sales of the originals, and avoid RIAA jumping in to kill
this new creative art form. It avoids DRM issues, and nicely
finesses the "Fair Use" of User's legal copies.
This would require creation of two programs: The instruction
generator "Compiler", which is used by the MashUp creator. And, the
MashUp Assembler engine which would use the instructions to generate
the desired MashUp. The instructions would actually be quite simple,
involving, selection from the original, and then either straight
copying, or blending/mixing with other selections (with optional
volume, pitch?, speed?) adjustments. If needed MD5 signatures of the
Song originals could be used to verify that the User has the
"Correct" version of the song, though this might not always work.
Now, given that we have these two applications the Instruction
Generator (Compiler) and Assembler. This same idea could be used to
reduce the bandwidth usage of certain style podcasts, such as yours!
My assumption is that Comcast and other broadband suppliers do
some transparent caching of content to avoid always re-fetching
common items. And also, that a dedicated listener would have several
pre-existing podcasts already on their computer.
Since the "Daily SourceCode" is composed of various bits,
bumpers, promos. Extend the RSS enclosure to add an optional
<Assembler> section. This would build the new podcast from smaller
segments, either fetched from the internet (which would hopefully be
cached at the broadband suppliers head-end), or optionally as
segments copied from known offsets in pre-existing podcasts.
I am assuming, that you are using CastBlaster or MegaSeg, with
pre-loaded sounds bits to play. I haven't done any analysis, but
your shows do have repeated segments. From something as simple as
Boing, to the PodShow bumpers, and promos. Now, sometimes you do
talk over this segments, (in which case maybe you only send the voice
track to mix into the pre-existing segment)
This PodCast Assembler idea, all falls apart with something like
Dawn and Drew, where there are very few repeated segments.
I realize this does a software effort, and also working with the
MashUp creative people. I am hoping that CastBlaster's general front-
end could be re-used, but replace the MP3 output, with a MP3
Assembler instruction generation module.
So, What do you think?
Craig
Ps. Simplified Example of Assembler instructions
Assuming song1.mp3, song2. mp3 and song3.mp3
<COPY FILE="song1.mp3" START="00:10:02.1" END="00:10:25.2"
NAME="background">
<COPY FILE="song2.mp3" START="00:06:02.0" END="00:06:12.3"
NAME="Chorus">
<COPY FILE="song3.mp3" START="00:00.0" END="00:00:23.1"
NAME="Vocal" PITCHBEND=".12">
<MERGE>
<NAME="background" Volume="0.5">
<NAME="Chorus" Loop="yes">
<NAME="Vocal">
</MERGE>
The Result would mix 23.1 seconds of song1 (at 1/2 volume) with a
looped 10 seconds of song2, and with 23.1 seconds of song3 (Pitch
increased by 12%)
or for the Daily SourceCode, something like:
<INSTRUCTIONS>
<SEGMENT_LIST>
<SEGMENT NAME="Vocal" SRC="dailysourcecode.com—mp3
DSC-2005-229-08-23-vocal.mp3">
<SEGMENT NAME="Rachmaninoff" SRC="dailysource.code
mp3/rachmaninoff.mp3" ALTERNATE="DSC-2005-228-08-22.mp3,
00:00:00.0-00:00:10.0">
<SEGMENT NAME="Bong" SRC="...." ALTERNATE="...">
<SEGMENT NAME="WalkTheDog" SRC="http://...."
ALTERNATE=".....">
<SEGMENT NAME="MashUp1" SRC="http://
viprhealthcare.typepad.com/mashup_of_the_week_podcas/files/
MJs_Under_Pressure.mp3">
</SEGMENT_LIST>
<CONCATENATE>
<MERGE>
<NAME="Rachmaninoff Opener" >
<NAME="Vocal" START="00:00:00.0" END="00.00.12.1">
</MERGE>
<NAME="MashUp1">
<NAME="Vocal" START="00.00.12.2" END="00.03.00.0">
.
.
.
</CONCATENATE>
</INSTRUCTIONS>
|
|
