17:59:51 #startmeeting Weekly Video Team Meeting 17:59:51 Meeting started Thu Aug 20 17:59:51 2020 UTC. The chair is tumbleweed. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:59:51 Useful Commands: #action #agreed #help #info #idea #link #topic. 18:00:02 #link http://deb.li/oNCD Agenda 18:00:05 #topic roll call 18:00:09 hollah 18:00:15 o/ 18:00:16 ohai 18:00:24 So. This is it. The last meeting before we're too busy for meetings 18:00:26 hey 18:00:26 * pollo waves 18:00:41 * wouter waves, too 18:01:15 #topic Video Stack - Jibri (passwords) 18:01:17 nothnig happening there 18:01:24 it's been done? 18:01:27 nope 18:01:32 Yes it is 18:01:35 oh, right, stream passwords were done, yes 18:01:41 password support; nothing happening there 18:01:56 unclear agendas++ (especially when I wrote them) 18:02:02 password support for rooms - no 18:02:07 o/ 18:02:17 #info Streaming is now password protected 18:02:19 o/ 18:02:21 paddatrapper: I assume you've tested it? 18:02:25 tumbleweed:yup 18:02:39 #topic Video Stack - Voctoweb 18:03:14 Still to do: group auth, notification of what other people are doing, some layout presets 18:03:25 at least, those are things I think are somewhat important, and time is getting tight 18:03:43 added VU meters to the previews. I'm not very happy with them, but they seem better than nothing 18:04:24 * pollo checks something,please don't change topic 18:04:25 oh, and I wanted to do playback improvements 18:04:38 seeking, playing from an offset 18:04:46 and interpreting wouter's metadata exports 18:04:51 yes, that's what I was looking at 18:04:59 imo seeking is the #1 priority 18:05:05 it may be hard, though 18:05:08 tumbleweed: ah yes, have you been able to look at those? 18:05:10 but I'll give it a try 18:05:11 wouter: nope 18:05:24 if you don't have time, what would be nice is the ability to pause and restart at the same time 18:05:29 tumbleweed: the metadata export should be deployed currently, but I haven't tested them yet 18:05:32 s/time/ place in the video/ 18:05:36 pollo: I think that's the same problem 18:05:41 hmm 18:05:57 I'm 100% we're going to have problems with that 18:06:08 "/event/2/bystate/preview" should work 18:06:18 sorry 18:06:27 if we can't get seeking, we need a fallback 18:06:29 https://review.online.debconf.org/api/v1/event/2/bystate/preview 18:07:17 except it needs an API key which isn't configured. I guess I can drop that for the time being. 18:07:24 tumbleweed: what could also be done is to remove the stop button 18:07:49 so you can start it, play another file if needed, but once it's playing you can't accidentally stop it 18:08:18 yeah, I wanted to put stop behind an Are you Sure modal 18:08:27 yeah - I am wondering if there is a need for pause 18:08:30 tumbleweed: I was just about to suggest that :) 18:08:50 I can't think of any reason to abort a playback, other than "oops, didn't want to do that" 18:08:55 pollo: I can imagine a fallback with ffmpeg, if nothing else 18:09:18 s/ffmpeg/ingest ? 18:09:56 ah, no we'll get the same problems with ingest 18:10:28 no, I meant ffmpeg 18:10:36 I'd be happy with a stupid script you'd call like "playback_fallback.sh video.mp4 00:02:23" 18:10:46 I'll bet you can pass in a start parameter, and if you type that in from the UI, you can guess at the number of seconds needed 18:10:50 anyway, out of the things I described 18:10:55 I can have a look at the non-playback voctoweb stuff 18:10:55 that seems like the priority, right? 18:11:02 tonight 18:11:08 \0/ 18:11:32 olasd: what sort of things are you thinking? 18:11:33 I would say picking a video to play is higher than pause/skip 18:11:36 unless I owe you something more urgent (but I guess we'll find out during the rest of the meeting) 18:11:39 CarlFK: we have that 18:11:50 #action pollo to write a ffmpeg wrapper for video fallback 18:11:51 oh. never mind. 18:11:52 tumbleweed: the layout presets and notifications? I guess 18:12:03 olasd: great :) 18:12:08 contributions welcomed 18:12:10 well, any of the pending stuff really 18:12:15 and I'll happily work with anyone to get them set up for local dev 18:12:41 we can look at that post-meeting, yeah 18:13:02 I'd say I'd help out too, except that I have enough on my plate as is :) 18:13:02 OK, my priority will be seeking 18:13:09 let's move on 18:13:18 #topic Video Stack - Loop Generation 18:13:30 I think highvoltage said he'd made some progress? 18:13:33 highvoltage: I see you have a prototype loop 18:13:44 o yeah finally some progress 18:13:56 new test loop as of a few hours ago: https://peertube.debian.social/videos/watch/3841db68-7344-4134-b7d2-ed4d9a23b1bf 18:13:56 can you explain a little bit how it works? 18:14:05 say I have to change something, what do I do? 18:14:13 since then I also added some dc17, dc18 clips 18:14:42 pollo: right now you can't, since I can't access vnc on obs1 for some reason. was troubleshooting before the meeting started and will continue after 18:14:51 but... let me talk a little about that anyway 18:15:23 there's a sponsor loop scene in obs, that automatically progresses to a "happening now" scene 18:16:20 whenever we land on "happening now" for whatever reason, it starts to play one of the videos in the random pool (which contains the shoutouts, memories from previous debconfs, did-you-know slides, and whatever else we might want to display periodically 18:16:24 ) 18:16:41 so, if you'd need to modify something, you'd need to vnc in to obs1 18:17:18 the tool that manages the auto slide changes is called Advanced Scene Switcher, it's a 3rd party tool 18:17:42 the plugin that displays the web content is called something like "Linux browser plugin" 18:18:04 I'll write up a bit more about both, a bit longer term I'd like to replace the scene changing logic with a shell script 18:18:16 so... Have we figured out how we're getting it into voctomix, yet? 18:18:18 that's it in a nutshell for now, we can delve into a it a bit later on 18:18:25 sounds like something we'll want to keep for further confs too 18:18:27 looks really nice though, I like it :) 18:18:31 yeah, agreed 18:18:46 obs needing a GPU is a bummer though 18:18:56 but the "get it into voctomix" bit is kinda important, how do you see that happening? 18:18:57 it doesn't actually 18:19:07 tumbleweed: I checked OBS, it is possible to set the raw matroska and a custom uri to pipe it too, so it might just work, it's one of the first things I'll get in to once I can actually get access to obs1 18:19:07 it works without the GPU, if you have a honking great CPU 18:19:52 ah, so that was mocked up at home 18:19:57 also, I'm going to take that sponsors clip and make an old fasioned kind of sponors loop from it just in case all of this doesn't end up working out 18:20:11 or well, ends up not working in time, which maybe slightly more likely 18:20:29 in the worstest case, we can stream the loop to an rtmp endpoint and inject that back into vocto with ffmpeg or gst 18:20:31 yeah, backup plans might help 18:20:37 olasd: yeah 18:20:40 although I think by tomorrow the kinks can be sorted out which leaves a little time for a little testing but it's all very tight 18:20:51 highvoltage: if you don't get tigervnc working, remote-viewer vnc://obs1.online.debconf.org:5900 should "just work" 18:21:07 as it downgrades to no verifying the TLS cert by default 18:21:30 pollo: it prompts for a password which I don't have 18:21:38 they will all do that :) 18:21:46 didn't I give you the password? 18:21:50 nope 18:21:57 I remember giving it to you on matrix 18:22:04 ah that rings a bell 18:22:15 I remember you both saying the password was OTR encrypted on matrix 18:22:18 too 18:22:41 that's probably just what I was missing for tigervnc too then. ok I think we can move on from that for now 18:22:56 heh 18:23:16 So, what are we going to do with the loop 18:23:19 at least that sounds like an easy fix 18:23:22 we've currently got a lopo source in voctomix 18:23:25 (yay so at least remote-viewer works with the password) 18:23:28 and we've got the loop blanking 18:23:44 we were thinking we may want to PinP the loop 18:23:53 can OBS stream to two endpoints? 18:24:07 well, my question was: Does anyone still think we need that? 18:24:09 atm the loop source and the loop blanking are 2 differents ports 18:24:37 I guess not 18:24:39 you're misnaming things which makes it confusing; there's two blanking inputs: an empty one and a loop one 18:24:56 olasd: in voctomix terms they're both blankers 18:25:00 IIRC 18:25:01 yes 18:25:23 but the point is that we currently have 2 loops being injected into voctomix 18:25:27 the stream blanking loop 18:25:29 and the source called loop 18:25:35 I see 18:26:13 the regular input is more versatile, and we can emulate the "stream blanking" feature with a preset 18:26:20 once they happen 18:26:26 yes 18:26:32 (switch audio + source a to loop, full screen) 18:26:32 which is why we did it 18:26:55 but what do we put on the stream blank source then? 18:26:55 ah yes, audio 18:27:00 nothing? 18:27:09 and disable the UI for it? 18:27:21 I don't think we're actually gaining any of the versitility then 18:27:36 and we've lost the ability to protect ourselves from recording screwups where we stay on blank 18:27:39 I'm not following, why do we even need an empty blanker? 18:27:42 I guess that won't be a problem with an online debconf, anyway 18:27:57 wouter: the UI for the empty blanker is already disabled 18:28:48 tumbleweed: the "loop preset" wouldn't preclude the loop input from being used in other layouts 18:29:11 as long as the controls are available, yes 18:29:16 yeah, sure 18:29:24 we should probably give highvoltage just one port to stream to 18:29:39 sounds easier yes 18:29:59 ok, sounds like the loop source is it then 18:30:03 and the other UI will have to be disabled 18:30:05 highvoltage: for the video shoutouts, how is audio handled? 18:30:17 let's use the regular input for now, we can switch that out if presets don't happen 18:30:45 I can see how widely fluctuating audio levels from shoutout to shoutout would be a problem 18:30:46 man, I wish we'd had this weeks ago 18:30:47 pollo: the audio is inside the video clip, the music is an embedded scene shared across the other scenes, but isn't contained in the shoutout scenes 18:31:23 let's move on 18:31:25 #topic Video Stack - Grabber 18:31:28 pollo: ah right, I can normalize a bit when editing the final shoutout clip in kdenlive 18:31:29 I think what pollo meant was "please make sure the audio from shoutouts is normalized" 18:31:33 Nothing new to say here, probably 18:31:33 pollo: but that's just going by ear 18:31:44 :) 18:32:04 highvoltage: SReview::Normalizer might help too, but I'll leave it up to you as to whether to use that... 18:32:07 tumbleweed: let's move on 18:32:08 #topic Video Stack - SReview 18:32:15 so, yes 18:32:24 wouter: is that just a setting? if so I'm all for it 18:32:29 wouter: does that mean you're volunteering sreview for people to upload shoutouts? 18:32:33 ... I'll have to re-encode everything again :-/ 18:32:42 :) 18:32:44 highvoltage: no, it's a perl module 18:32:55 oh nm then 18:32:57 highvoltage: API is fairly simple 18:33:36 highvoltage: SReview::Normalizer->new(input => SReview::Video->new(url => "/path/to/input"), output => SReview::Video->new(url => "/path/to/output"))->run() 18:33:37 yeah but I have nowhere to put that in (qotd) 18:33:41 heh :) 18:34:21 I don't want to add another upload script though, so if you don't want to use it... your issue :) 18:34:23 is there more to report about sreview? 18:34:35 a bit 18:34:48 uploads seem successful, mostly, except for bug that I just pushed the fix for 18:35:27 I did write the script to sync states between the three events based on what's uploaded etc, but when I ran it the first time today I saw there were some issues with that still 18:35:55 I'll fix that either today or tomorrow, not sure yet 18:36:11 beyond that, I only need intro/outro graphics 18:36:14 which I haven't seen yet 18:36:45 (I'll retranscode things after the meeting, don't want to KGB-bomb the meeting right now) 18:37:14 #info we need to come up with intro/outro graphics for talks 18:37:15 that's it 18:37:24 it's probably time to start making some fall-back graphics 18:37:30 you mean title cards? 18:37:33 yeah 18:37:33 or something fancier? 18:37:35 nattie: those, yes 18:37:35 ah 18:37:38 yeah, i need those too 18:38:01 https://salsa.debian.org/debconf-team/public/data/dc20-online/-/issues/11 has some info 18:38:15 but nobody did anything for those 18:38:31 if worst comes to worst, I can look at it myself, but then it will delay things 18:38:49 at any rate, I don't think this is a huge priority if there are other things 18:39:04 wouter: are the recordings going to have cards? 18:39:05 we can delay transcoding talks for a few days if there are more urgent matters, won't be a problem 18:39:31 tumbleweed: the ones you'll play in voctoweb you mean? No, not from SReview, different path through the system 18:39:45 OK, so we only need them for publishing final videos 18:39:52 exactly 18:40:15 and we can publish final videos whenever, I'd say 18:40:21 (well, not three months after, but you get the idea) 18:40:27 yeah 18:40:36 would be nice to have those in place before the conference starts, though 18:40:39 well we /can/; not saying that we /should/ :D 18:40:43 if possible, sure 18:40:58 let's move on 18:41:00 if people don't have time to get things ready in time because other, more urgent, matters take priority, then it's not a train smash 18:41:03 sure 18:41:15 I don'tknow if that's the issue here or not 18:41:23 jeff3rson[m] hasn't been replying on the topic 18:41:26 I don't know what the issue is :) 18:41:31 we'll see what happens 18:41:55 #topic Video Stack - Streaming 18:42:23 So, upgraded the digital ocean instances a bit, to get more transit quota 18:42:34 sorry for dropping all the balls here 18:42:35 what does it have now? 18:42:39 in their pricing model, you earn quota by the number of hours your instances are alive 18:43:04 but their transit pricing is also an order of magnitude cheaper than any other cloud provider, so paying overages isn't a big deal there 18:43:20 highvoltage: the current instances are the $20/month ones 18:43:38 no idea if those are big enough for our load or not 18:43:39 and what kind of throughput do we get on those? (assuming it's ever been tested) 18:43:45 but they're the sweet-spot for bunled quota 18:43:50 it hasn't 18:44:05 we should try to load test 18:44:16 e.g. by playing something back and asking #debian-devel to watch or something 18:44:30 at least it sounds easy to just spin up more droplets in a worst case scenario 18:44:34 hackers? :P 18:44:44 highvoltage: yeah, relatively 18:44:49 paddatrapper: we'd get copyright issues if we did that publically, I'm afraid 18:44:50 * olasd DMCAs paddatrapper 18:44:50 paddatrapper: good way to get a DMCA takedown 18:45:00 but we could do the duck video thing 18:45:09 although our geoip distribution doesn't support round-robin amongst options, yet 18:45:12 it'd need to be added 18:45:22 lol, DMCA is a walk in the park, we have th FPB 18:45:25 play back on old debconf talk 18:45:32 wouter: hence the ':P' 18:45:37 heh 18:45:43 anyone want to take the action to do a test? 18:46:11 Sure 18:46:32 #agreed paddatrapper to organise a public video streaming load test 18:46:38 next question on this topic 18:46:43 who is going to be on call for the opening? 18:46:51 it's going to be 3am for me, so I'll be fast asleep 18:47:03 we should start trying to fill those "on call" slots 18:47:07 don't worry, we'll page you if there's an issue 18:47:10 :D 18:47:16 * wouter looks at the schedule 18:47:16 I think that first morning is going to be the critical time 18:47:24 I'll be there but I don't know how anything works 18:47:25 after day 2, video team life is usually fairly easy 18:47:36 until we start changing everything 18:47:41 heh 18:47:58 I can be there, would be 12:00 here 18:48:15 I'm unavailable for the first block of talks on Sunday unfortunately 18:48:21 whoever is there should probably get familiar with the streaming stack 18:48:34 tumbleweed: we have docs, right? 18:48:38 *ish 18:48:38 some 18:48:41 I'll have a play with things on saturday 18:48:55 I'll be there too 18:49:02 we're lacking on docs around the internal links between everything 18:49:05 I can also bother paddatrapper by phone if things really break down ;-P 18:49:11 wouter: note that we'll have training on sat, so please don't mess up thing during training :) 18:49:21 pollo: perhaps I should join the training, then 18:49:40 wouter: I'll be in the middle the sea in Hout Bay, so won't really be able to answer... 18:49:53 olasd wrote a lot of the streaming stuff, so having him there is probably the most valuable thing :) 18:49:55 paddatrapper: ooohh, Hout Bay, fancy 18:50:04 s/fancy/cold :P 18:50:05 isn't daytrip Wednesday? 18:50:08 that too 18:50:25 let's move on 18:50:26 Training schedule: https://wiki.debian.org/DebConf/20/VideoVolunteering 18:50:35 #topic Video Stack - Etherpad 18:50:37 Nothing new here 18:50:42 again, it probably needs group auth 18:50:55 not that complex, just haven't done it 18:51:39 #topic Video Stack - Embed in Website 18:51:47 See: https://debconf20.debconf.org/schedule/venue/1/ 18:52:13 there'll also be the usual stuff at the top of the front page once things start 18:52:25 I do wonder if we need a "how to attend debconf" document and email 18:52:47 it's a bit weird to have the video sandwiched between the title and the abstract of the next session 18:52:56 tumbleweed: I and nattie started that 18:53:01 https://pad.online.debconf.org/p/announcement-mails 18:53:01 fwiw I'll add that matrix bridge for #dc20-talks in the meantime too 18:53:15 tumbleweed: as a badger or as an mail to -announce? 18:53:32 ah, Isee you added that great 18:53:34 nattie: either works 18:53:42 (and btw we now have a matrix instance on matrix.debian.social that you can also access via the element web interface on element.debian.social, auth via salsa login) 18:53:56 which is probably easier to use than IRC for newbies 18:54:15 great @ matrix 18:54:36 olasd: I guess MRs welcome on that page :) 18:54:43 that video view page is really nifty 18:54:47 tumbleweed: I knew you'd say that 18:54:49 :P 18:55:13 it's working, that what matters 18:55:23 I see you fixed the quality selector 18:55:32 (generic you) 18:55:34 yep 18:55:38 and it's a big improvement from the debconf video docs page 18:56:13 (and somehow the video starts much faster, maybe it's my imagination, or maybe this is the stream that doesn't use hls anymore?) 18:56:27 it's HLS 18:56:30 /win 20 18:56:31 it's the same player hitting the same backend 18:56:50 maybe the one on debconf.org is more up to date 18:56:50 ah all in the mind then :) 18:56:59 it is a little more up to date, yes 18:57:27 moving on 18:57:29 * highvoltage mutes olasd and pollo that just started yelling on my speakers 18:57:35 did someone just switch the stream to something else? 18:57:36 I was about to say the same 18:57:45 #topic Streaming setup 18:57:47 * pollo ducks 18:57:47 err I think we did that 18:57:57 mini-dc marseille 18:58:00 #topic Advice/training for directors 18:58:11 training is on Saturday, whee! 18:58:12 paddatrapper! 18:58:24 Signup and schedule: https://wiki.debian.org/DebConf/20/VideoVolunteering 18:58:26 not many people have signed up yet 18:58:29 apparently 18:58:31 the email hasn't been sent yet 18:58:43 oh, that would explain that :) 18:58:52 the necessary changes have landed, just need to deploy them 18:58:54 #info training schedule up, email to be sent out 18:59:14 shouldn't wait too long with that I guess, it's in two days 18:59:16 * tumbleweed deploys 18:59:23 yeah, we should get that mail out today 19:00:30 nothnig more there, I guess 19:00:50 #topic Advice/training for presenters 19:01:01 I think there's some going on rgiht now? 19:01:05 we're doing test calls 19:01:19 though possibly we need to send out another badger reminding people to stop by 19:02:34 anything else on that subject? 19:02:59 how's talk feedback going? 19:03:22 urbec: you were doing quite a bit of review, what are they like? 19:03:54 there were a few issues with broken audio that were my fault 19:04:09 (which will hopefully be fixed after my re-transcode) 19:04:29 I found them mostly ok 19:04:33 From what I could gather they looked good otherwise 19:04:34 did that includ vagrant's one? from what I recall that was one that urbec mentioned that had quite bad audio 19:04:40 I mean the A/V sync issues 19:04:44 I did and talked to vagrant 19:04:47 those were my fault 19:04:55 no time to re-record, no better mic 19:04:59 wouter: please also check what's wrong woth maddog's talks (404) 19:05:11 terceiro: I did, already found the problem 19:05:15 i guess we'll just have to accept the bad sound 19:05:26 terceiro: I'll have to re-submit everything, which I'll do after the meeting 19:05:27 for which i apologise, since i mentioned it was bad, but then waved it through 19:05:40 wouter: cool, thanks 19:05:41 but a few (see backlog) 19:06:14 OK so it seesm things are reasonable 19:06:24 pity we coludn't get the bad ones fixed in time :( 19:07:17 #topic Salsa issues 19:07:24 let's skip over that 19:07:29 #topic actions from last meeting 19:08:43 paddatrapper: you tested simultaneous jibri streaming? 19:08:46 not sure how we would even 19:08:53 tumbleweed: yup and it works 19:08:55 cool 19:09:04 the new stream replaces the old cleanly in Vocto 19:09:21 the use case is someone closing the old jitsi room before stopping the stream 19:09:26 what we'd expect, great 19:09:31 And we need to get the next one up ASAP 19:09:50 next one? 19:09:59 next talk 19:10:06 ah 19:10:12 the streaming + irc + etherpad page is done 19:10:15 streaming passwords are done 19:10:25 test calls are happening (as of now, at least) 19:10:55 and paddatrapper did some writeup on the tools and architecture (at least as part of the volunteer docs) 19:11:19 I assume the olasd items keep carrying forward 19:11:25 Yeah not sure if we should go more in depth there 19:11:42 https://debconf-video-team.pages.debian.net/docs/room_setup.html#online-conference 19:11:55 I'll try to read through as much of those docs as possible when I've gotten to the minimal other responsibility dependencies 19:12:27 #topic AoB 19:12:57 should we maybe have another meeting saturday night 19:12:58 the minidebconf in regensburg has been postponed 19:12:59 ? 19:13:15 it's not happening at the end of september 19:13:26 grr volunteer page 500s if you're not logged in 19:13:34 I keep forgetting about anonymous users 19:13:43 saturday night would be our last chance to fix issues if there are any 19:13:58 (and with "night" I mean 18:00 UTC) 19:14:13 so a brief check-in on saturday? 19:14:35