Permalink
Switch branches/tags
Nothing to show
Find file
Fetching contributors…
Cannot retrieve contributors at this time
176 lines (128 sloc) 5.55 KB
config management
do this better: veyepar/setup/nodes/encode/push_confs.sh
anisebel?
veyepar:
talk setup checklist-
confirm talk title
ask presenter how they prounounce their name
ask audiance for a remote monitor (for IRC)
move state data def from models.py to initial_data.json fixture.
use state table, not a constant
episode.html save..
UI to show user it got saved.
"saved" page, 1 sec redirect back to episode?
js to say "saved" that goes away in 5 seconds?
js to say "saved/dirty" if there are outstanding edits?
When an Edit is saved, move to next_episode
next/prev (match recording sheet: same room/day)
# split long talk (tutorials) into parts.
error handling:
encode and uploads nead attention
log error, continue with next episode.
encode errors:
inkscape svg/png fail
input files missing
encoder fail
command line parameter to set cfg file.
option to disable reading config (tests should not read config)
give test script its own cfg
upload category -
needs default, currently hardcoded to "Education"
youtube:
playlists - 1 for show. maybe ndv, presenter, tags?
use testing server for testing
https://developers.google.com/youtube/2.0/developers_guide_protocol_testing
don't re-upload the same video (causes duplicate video error)
# invalid tags cause fail and loop forever - fix this.
(fixed by not setting tags)
state - need something to handle Canceled talks, breaks, etc.
disposition?
priority?
state - mk_titles, email_titles ?
(currently are done outside of the formal workflow steps, so hard to fix one and re-do)
state - really needs something new. like a bunch of "completed" flags.
flag - do not overwrite when syncing with conf db -
this is to handle when I know better than the conference data.
tweet hashtag - client or show or both?
pyvideo.org category - client or show or both?
(are these two the same thing?)
richard:
reference back to veyepar page (hidden to public)
reference to conference site page (for slides)
move show.schedule_url to client.
Stop - make it work again -
I think it is a buffering problem -
the code needs to poll the db, not look at the local cashed value
zip title svg and images
http://bazaar.launchpad.net/~inkscape.dev/inkscape/trunk/view/head:/share/extensions/svg_and_media_zip_output.py#L208
meh, just include all the images in the svg.
encoding:
remove global -dv-format setting
encode shorter talks first so they can get started uploading sooner.
soften jump cuts - If end+start is not back to back, fade to black and back up.
slugify talk titles-
truncate on word boundary when over X chars.
model:
episode:
child table of destination hosts (youtube, archive.org, mirror...)
child table of output files
mp4, ogv, mp3, facc...
text: irc logs, transcript, tweets
md5sum of file, and or that other thing that rackspace cdn uses
datetime encoded
grandchild of uploaded to urls?
datetime uploaded
track when deleted?
Once uploaded to a file sharing host, (archive, rax, mirror...)
use a cloud process to read from it and write to youtube so
that the video file only goes across local connection once.
option to pick transport: popen(rsync...) or rax_uploader
manage cloud storage:
png/svg for preview
ogv/webm for remote editing
audio_image for remote diagnostics
final encode for archive
process to check if a file is avalible for public viewing:
(neither of these is a real problem, so low priority)
youtube - "still processing, check again later."
rsync or such, conjob or human processes that veyepar doesn't know about
change duration from "hh:mm:ss" to float(sec)
BucketID / Richard Category
Sometimes this is per client, sometimes per show.
thought: one field in each, if Show else look to client.
wallpaper
logos:
NextDayVideo,
Linux, DVswitch, Python, Django, MLT
recording sheets:
hostname:____
add date/time to "Replace batteries"
add start-end time to header
(to help know how long the day is,
and help keep track of what the last sheet is.)
for unit tests:
get sphinx or some such speach recognition packaged and in a PPA
sphinx2-bin fell out of the ubuntu repos due to not being maintained.
state of the art is not a requirement
it just needs to recognize a simple phrase, like
"go forward 10 meters"
which gets encoded and demuxed back out
to make sure the encoder didn't muck it up.
Transcriptions
Need a place to host transcriptions
mix transcriptions with video for sub titling
aurora .. um.. that's all I got.
OCR - http://software-carpentry.org
flag for odd audio (like missing or very low)
DD - "There is a new audiolevel filter. See output of melt -query filter=audiolevel"
ponder and debate cred management.
voluneer management.
flight details: bday, window, ff#
list on credit slide.
(veyepar) carl@twist:~/src/veyepar/dj/scripts$ python sync_rax.py --raw --low
[mpeg1video @ 0x7fdc340cb7e0] interlacing not supported by codec
[mp2 @ 0x7fdc342935e0] Channel layout not specified
[mpeg @ 0x7fdc340008c0] VBV buffer size not set, using default size of 130KB
If you want the mpeg file to be compliant to some specification
Like DVD, VCD or others, make sure you set the correct buffer size
[swscaler @ 0x7fdc106311a0] Warning: data is not aligned! This can lead to a speedloss