Friday, 2017-01-20

*** tpb has joined #timvideos00:00
*** tpb has joined #timvideos00:08
xfxfhttp://veyepar.nextdayvideo.com/main/C/lca/S/lca2017/E/12058/00:57
tpbTitle: veyepar: Rewriting MMU for fun and profit (at veyepar.nextdayvideo.com)00:57
CarlFKmelt -profile hdv_720_30p /home/videoteam/Videos/veyepar/lca/lca2017/mlt/Lightweight_inventory_management_with_Pallet_Jack.mlt [email protected]/45 -consumer avformat:/home/videoteam/Videos/veyepar/lca/lca2017/webm/Lightweight_inventory_management_with_Pallet_Jack.webm progress=1 threads=0 ab=256k vb=2000k quality=good deadline=good deinterlace=1 deinterlace_method=yadif01:06
tumbleweed30p sounds wrong :)01:06
xfxf^ that01:07
CarlFKHmm01:07
xfxfalso hdv seems wrong too, and you don't need to deinterlace, it's already progressive01:07
tumbleweedis xfxf freaked aout about that, yet?01:07
tumbleweedls /usr/local/Shotcut/Shotcut.app/share/mlt/profiles/01:08
tumbleweedhdv_720_25p ?01:09
tumbleweedthat seems to be what we recorded01:10
xfxfall the existing files have been using the hdv_720_30p preview :/01:10
xfxfbut whatever01:10
tumbleweedthat's the bit to (not) freak about01:10
xfxfthings are in sync, most people aren't going to notice/care01:11
tumbleweedit means end cuts are early which is why CarlFK is looking at this01:11
xfxfooohh01:11
xfxfoh, wait, wat01:11
xfxfthis might explain why i ketp sticking some videos back into the encode queue because they didn't finish at the proper time01:11
xfxfbut everything i approved seemed to end correctly01:11
tumbleweedok01:12
xfxfi'm double checking now01:12
tumbleweedmay still be worth re-doing, if there's horrible frame rate conversion01:12
tumbleweedwe have the CPU horsepower, I think01:12
tumbleweedI think I haven't hit this before, because wouter wrote us a completely separate encoding script (lol, sadface)01:13
CarlFKI am pretty sure wouters hack re-encoded the mp4 that veyepar/melt created.01:13
tumbleweedlol01:14
xfxfjust checked 3 talks, all end at the right time01:14
CarlFKhe just wanted to adust the audio levels.01:14
xfxfoh01:14
xfxfi remember what I usually do01:14
xfxfi extended your enc script to read in a second set of upload_formats01:14
xfxfand added a new state01:14
tumbleweedCarlFK: well, we also produce webms, not mp4s01:14
xfxfso i used whatever melt webm stuff was already there01:14
xfxflet me see if I can find my older veyepar install01:15
CarlFKCurrent Frame:      17473, percentage:         7501:16
CarlFKthat is makeing a webm and going about 1 frame every 2 seconds.01:16
CarlFKCurrent Frame:      17483, percentage:         7501:16
tumbleweedCarlFK: what machine? that's crazy slow01:17
CarlFK30 seconds for 10 frames01:17
CarlFKenc6 (bigbox right?)01:17
tumbleweedyeah01:17
CarlFKit only uses 1 core.  but still...01:17
CarlFKmelt -profile hdv_720_30p /home/videoteam/Videos/veyepar/lca/lca2017/mlt/Lightweight_inventory_management_with_Pallet_Jack.mlt [email protected]/45 -consumer avformat:/home/videoteam/Videos/veyepar/lca/lca2017/webm/Lightweight_inventory_management_with_Pallet_Jack.webm progress=1 threads=0 ab=256k vb=2000k quality=good deadline=good deinterlace=1 deinterlace_method=yadif01:17
tumbleweedwhatever it's doing seems single threaded01:18
tumbleweedno need to deinterlace01:18
CarlFKpretty sure the encoder is single threaded.  threads=0  means "figure it out"01:18
tumbleweedyeah01:18
tumbleweedI seem to remember that the vp8 encoder is01:18
tumbleweednumber of threads to use for encoding, can't be 0 [auto] with VP801:19
tumbleweedwell there you go :P01:19
xfxfok, just checked what i last used 6 months ago, it's the same01:19
CarlFKwhat does "can't be 0" mean if I use 0?01:19
CarlFKor should I just follow directions and not use 001:19
tumbleweedprobably that it doesn't do any automatic guessing01:19
xfxfin veyepar.cfg01:19
xfxfupload_formats=mp401:19
xfxfsecond_upload_formats=webm01:19
xfxfdv_format=hdv_720_25p01:19
tumbleweedaha, we don't have that in our veyepar.cfg01:20
CarlFKI see a #commented out line:01:20
CarlFKcmds=["melt %s -profile dv_ntsc -consumer avformat:%s progress=1 acodec=libvorbis ab=128k ar=44100 vcodec=libvpx minrate=0 b=600k [email protected]/3 maxrate=1800k g=120 qmax=42 qmin=10"% (mlt_pathname,out_pathname,)]01:20
xfxfthat looks like it's for 4:3 DV01:20
xfxftumbleweed: yeah, second_upload_formats was a hack I added for a second enc.py01:21
CarlFKshould we pluck out the  codec's and anything else...01:21
tumbleweedCarlFK: I'm fine with threads=0 meaning 1 thread, if we run 12 parallel encoders on bigbox01:22
tumbleweedbut more practically, we probably want to bump it to 3 or something01:22
tumbleweedmaybe 4 would be rounder01:23
CarlFKright now 1h or vid will take 30 hours to encode01:23
tumbleweedxfxf: we could rope r6mix into encoding...01:23
tumbleweedyeah, that sounds all wrong01:23
tumbleweedkill the deint, specify encoders, and we should be able to get performance up01:24
xfxfwebm is definitely slower, the encoder is less optimised01:24
tumbleweednot that slow01:24
CarlFKxfxf: melt -verbose -progress -profile hdv_720_30p /home/videoteam/Videos/veyepar/lca/lca2017/mlt/Decoding_Satellites_with_SatNOGS.mlt -consumer avformat:/home/videoteam/Videos/veyepar/lca/lca2017/mp4/Decoding_Satellites_with_SatNOGS.mp4 threads=4 progressive=1 strict=-2 properties=x264-high ab=256k01:24
xfxfbut it's like 4-5x slower, not 50000x slower01:24
tumbleweedhdv_720_30p is still not correct, surely01:24
tumbleweed25p01:24
tumbleweedand you've got a ,mp4 in there01:25
tumbleweedthat's a typo01:25
tumbleweedoh, no, dirt on my monitor :P01:25
CarlFKum.. wut.. right.. hang on...01:25
CarlFKoh right.. that was the messed up ending, not my webm test01:26
CarlFKmy bad.. hold please..01:26
xfxfnp, i'll fiddle01:27
tumbleweedif either of you do veyepar.cfg fiddling, please tell me01:27
xfxfby fiddle i mean independent of veyepar01:28
xfxfi'll paste back a better set of params to feed melt01:28
CarlFKI added dv_format=hdv_720_25p to gw1 .../veyepar/dj/scripts/veyepar.cfg01:28
CarlFKLightweight_inventory_management_with_Pallet_Jack.webm01:29
CarlFKah,      sh.writelines(log_text); self.run_cmd(cmd); sh.close()01:32
CarlFKthat is why  don01:32
CarlFKthat is why  don't see the line that is still running01:33
CarlFKtumbleweed: rsyncing?01:33
CarlFKI am going to go to green room and get food01:33
xfxfmelt -verbose -profile hdv_720_25p /home/videoteam/Videos/veyepar/lca/lca2017/mlt/Decoding_Satellites_with_SatNOGS.mlt -consumer avformat:/tmp/test.webm progress=1 threads=4 ab=256k vb=2000k quality=good deadline=realtime cpu-used=001:54
micoloustumbleweed: tasman A is losing sync again02:31
tumbleweedmicolous: yays02:31
tumbleweedI see high WER02:32
tumbleweednew cable?02:32
micolousthe av people here are on the phone02:32
micolousI assume with someone else in noc02:32
micolousIt's come good again now02:33
tumbleweedWER is back to zero02:33
micolousIt glitched out again for about a second02:36
tumbleweeddvisampler0: ph:  22   18   18 // charsync:111 [3 3 3] // WER:5199   0   0 // chansync:1 // res:1280x002:36
tumbleweedthen back to 002:36
tumbleweedoh, high again02:36
micolousYeah, it's stuck bad02:36
tumbleweedwe need to change that cable, I think02:36
micolousI think tell AV to change cable02:36
micolousyeah02:36
tumbleweedour cable, not theirs02:36
jeai did `input0 on` earlier to fix the continuous problem02:36
tumbleweedbut yeah02:37
micolousThey're on the phone02:37
micolousDid they put in their own cable?02:37
tumbleweedI mean, this is our problem, not theirs02:37
micolousSpeaker is persisting02:38
micolousAs best they can02:38
jeai can't do input0 on continuously unfortunately02:38
tumbleweedis anyone in the noc doing anything about a new cable?02:38
jeathe last time i did it, nothing fixed it02:39
tumbleweednot that I'm sure it'll help02:39
tumbleweedbut Iv'eb ene threatening to replace it for a couple of days now02:39
jeai'm the only one in the noc at the moment. will a new cable actually work?02:39
tumbleweedit's worth a shot, if nothing else02:39
tumbleweedanother opsis is another option02:39
jeayeah, it goes bad as soon as i do input0 on02:39
jeai don't like interrupting talks, but maybe i need to for this on02:39
tumbleweedI'm going to come and do a cable swap02:39
jeaok02:39
jeait just went back to normal02:40
jeawithout me changing anything02:40
* tumbleweed has swapped the cable02:47
tumbleweedI put the opsis from wellington 2 on top of the one we're currently using02:47
tumbleweednext thing to try is the other opsis02:47
jeaok, thanks for that02:49
jeacable seems to have fixed a lot of it already, so hopefully that will be enough02:50
tumbleweedthat cable goes in the bin then02:50
tumbleweedor into tim's bag of "interesting" cables02:50
jeayeah, give it to tim02:52
jeai wonder if it was one of his new ones or not02:52
jeawas it a white one?02:52
tumbleweedblack02:52
jeaok02:52
tumbleweedthe two in room X are white02:52
CarlFKthere are black magic cables?02:52
tumbleweedCarlFK: yep, and red, and blue02:52
jeanow i'm trying to remember if tim's unopened ones were black or white02:53
tumbleweedblack 3ft I think02:53
jeatumbleweed: i've never heard of a bluemagic. where can i buy one?02:53
tumbleweedI think carl has one :)02:53
jeawhat outputs does it have?02:54
tumbleweedmagical ones?02:54
CarlFKhdmi to rj4502:54
xfxfmelt -verbose -profile hdv_720_25p /home/videoteam/Videos/veyepar/lca/lca2017/mlt/Decoding_Satellites_with_SatNOGS.mlt -consumer avformat:/tmp/test.webm progress=1 threads=4 ab=256k vb=2000k quality=good deadline=realtime cpu-used=0 vcodec=libvpx03:44
CarlFKwow.. my first test still not done04:11
CarlFK^Crrent Frame:      21608, percentage:         9304:11
CarlFKkilled04:11
CarlFKyeah, this works.  yay. go fast go.04:24
CarlFKrun time: 13.566666666666666 minutes04:29
*** ivodd has joined #timvideos07:54
*** ivodd_ has quit IRC07:59
*** CarlFK has quit IRC08:57
*** CarlFK has joined #timvideos14:19
*** ChanServ sets mode: +v CarlFK14:19
*** rohitksingh has joined #timvideos14:45
*** ivodd has quit IRC14:52
*** ivodd has joined #timvideos14:52
*** sb0 has joined #timvideos15:43
*** NTU has quit IRC15:45
*** NTU has joined #timvideos15:47
*** rohitksingh has quit IRC15:51
*** rohitksingh has joined #timvideos15:51
*** rohitksingh has quit IRC15:52
*** rohitksingh has joined #timvideos15:54
*** rohitksingh has quit IRC15:57
*** rohitksingh has joined #timvideos15:58
*** rohitksingh has joined #timvideos16:02
*** rohitksingh has quit IRC16:03
*** rohitksingh has joined #timvideos16:05
*** rohitksingh has quit IRC16:08
*** rohitksingh has joined #timvideos16:13
*** rohitksingh has quit IRC16:23
*** rohitksingh has joined #timvideos16:24
*** rohitksingh has joined #timvideos16:28
*** rohitksingh has quit IRC16:36
*** rohitksingh has joined #timvideos16:38
*** karan has joined #timvideos16:55
karanmithro: ping16:58
*** sb0 has quit IRC17:19
*** karan has quit IRC17:36
*** sb0 has joined #timvideos17:54
*** sb0 has quit IRC17:58
*** sb0 has joined #timvideos18:20
*** sb0 has quit IRC18:22
*** sb0 has joined #timvideos18:44
*** sb0 has quit IRC18:45
*** sb0 has joined #timvideos19:08
CarlFKgrumble ;)  /me goes back to bef19:32
*** sb0 has joined #timvideos19:32
*** sb0 has quit IRC19:33
*** sb0 has joined #timvideos19:57
*** sb0 has joined #timvideos20:21
*** sb0 has joined #timvideos20:45
*** sb0 has quit IRC20:45
*** CarlFK has quit IRC23:22

Generated by irclog2html.py 2.13.1 by Marius Gedminas - find it at mg.pov.lt!