18:00:22 #startmeeting 18:00:22 Meeting started Thu Jul 16 18:00:22 2020 UTC. The chair is paddatrapper. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:22 Useful Commands: #action #agreed #help #info #idea #link #topic. 18:00:22 s/to// :) 18:00:33 Agenda: http://deb.li/oNCD 18:00:40 #topic rollcall 18:00:52 o/ 18:00:57 o/ 18:00:59 yo 18:01:00 * wouter throws away the trouts and waves 18:01:12 \o 18:01:20 I wondered who was smelling fishy 18:01:20 0/ 18:01:26 o/ 18:01:26 #topic Actions from last meeting 18:01:54 #info presenter advice has been merged, but will see a couple more tweaks today 18:02:28 what is the link to the previous meeting's notes again? 18:02:35 https://ssa.paddatrapper.com/debconf-video/2020/debconf-video.2020-07-09-18.00.html 18:03:01 #info olasd has set up RTMP streams with ~5s delay 18:03:04 thanks -- now that will show up in the current meeting logs, too ;) 18:03:18 #link https://ssa.paddatrapper.com/debconf-video/2020/debconf-video.2020-07-09-18.00.html 18:03:21 #info Jitsi has been ansibled 18:03:41 pollo: you tested audio sync right? 18:03:53 I think someone did, but I didn't 18:04:07 so, I did do a bit of SReview hacking to allow for uploading videos, but I need to redo a few things 18:04:14 I messed up and added some bad design 18:04:25 also, I need to test that the code works properly, haven't done that yet 18:04:28 but beyond that, should be fine 18:04:32 so, I got the geoip bit done. But didn't tackle anything VNC related 18:04:47 "CarlFK woo, look at the stream, and share my satisfaction of AV sync test goodness " 18:04:53 #info geopip script has been done 18:04:54 so I guess it works yes 18:05:08 #info AV sync in vocto with pre-recorded talks tested and working 18:05:31 olasd: did you manage to get a list of available recordings into voctoweb? 18:05:48 #action tumbleweed to poke VNC on vocto machine 18:05:49 I've been somewhat hacking at recording playback 18:05:49 #info SReview upload stuff in progress, something is there but needs more work 18:06:14 the list part is easy, getting the playing working is less easy (I'm trying to mix 2 asynchronous python frameworks...) 18:06:17 #action wouter to finish up SReview upload and test 18:06:26 yeah ok, that is tricky 18:06:33 shall I action you then? 18:06:45 paddatrapper: I haven't because someone recoded voctoweb before I could :P 18:06:51 heh true 18:07:00 (but really I've been distracted during the sprint) 18:07:22 paddatrapper: yeah, can do 18:07:27 I've got that ticket assigned to me, too 18:07:52 #action tumbleweed to add a list of available recordings to voctoweb and have them play on command 18:08:01 #action paddatrapper to investigate easy switching of Jitsi from 720p to 480p for large bofs 18:08:19 Think that was all the points from last meeting 18:08:22 Have I missed any? 18:08:33 tumbleweed: eh, SReview creates the videos in a particular directory, with filenames after the talk's slug 18:08:45 wouter: sounds good 18:08:46 tumbleweed: might it be easy to just mount that over NFS or some such from voctoweb? 18:08:55 probably rsync them over 18:08:59 or that 18:09:04 * tumbleweed doesn't want video streams depending on NFS 18:09:13 fairy nuff 18:09:32 #topic Blocked salsa issues 18:09:34 #link https://salsa.debian.org/debconf-team/public/data/dc20-online/-/issues 18:10:07 Anything in particular that needs discussion? 18:10:31 Do we want to have an "easy" mode in voctoweb before starting to work on the director guide? 18:11:18 might make sense 18:11:19 pollo: how would you see that working? 18:11:35 the current voctoweb has a bunch of buttons and no presets 18:11:35 it's easy to say "we need an easier version", it's not so easy to make the easier version 18:11:50 the idea was to hide some buttons and replace them by presets 18:11:52 I think the idea is to have a few set scenes 18:11:58 OIC 18:12:04 which control A+B source, layout, and audio 18:12:08 yeah, that could make sense 18:12:11 that's pretty straightforward to implement 18:12:19 Presenting scenes in the 18:12:22 I think ivodd is quite keen to hack on it 18:12:29 if we already have a plan for what to do, then fine, ignroe what I said ;) 18:12:31 (or for me to, not quite sure where we stand, there... :P) 18:12:35 *ignore 18:12:37 Presenting scenes in the 'not easy' version would be helpful too 18:13:31 one thing about voctoweb: audio control is going to be almost non-existant 18:13:37 the director can't hear the audio 18:13:41 and latency... 18:13:51 So, our sources will need to be well normalized 18:14:05 wouter: will sreview normalize the provided recordings? 18:14:09 #agreed Have a 'easy mode' in voctoweb that hides some of the buttons and gives the user presets to work with 18:14:12 tumbleweed: not currently 18:14:21 tumbleweed: I can add that, but it might introduce clipping too 18:14:26 at the very least we should manually review for normalization 18:14:40 I'm pretty sure we agreed on that last meeting 18:14:44 may be possible to normalize at ingest time 18:15:06 I'd prefer if audio normalisation was done in Sreview if possible 18:15:13 Or we can default to not normalizing, but having the option to normalize if it's necesary 18:15:15 one less thing to do on the fly 18:15:27 ah, meh, I'll just do it 18:15:39 probably going to need options to compress, too 18:15:49 if there's clipping we can see about fixing it later 18:15:58 apparent loudness is probably more important than normalization 18:16:04 we should try to get a baseline on jitsi and normalise on that 18:16:08 yeah 18:16:09 that's what bs1770gain does 18:16:13 #action wouter to look at normalising pre-recorded talks in SReview 18:16:17 apparent loudness 18:16:40 by normalizing -- yeah, that's a bit weird, but hey 18:17:05 DESCRIPTION 18:17:05 BS1770GAIN is a loudness scanner compliant with ITU-R BS.1770 and its 18:17:08 flavors EBU R128, ATSC A/85, and ReplayGain 2.0. It helps normalizing 18:17:11 the loudness of audio and video files to the same level. 18:17:26 +1 18:17:42 #topic Video stack 18:17:58 anyone have an opinion to kicking bad audio back to the presenter to do it better? 18:18:06 CarlFK: that's probably the ideal 18:18:07 I think we can officially say we'll not be using OBS 18:18:20 pollo: yeah, sounds reasonable 18:18:23 +1 18:18:36 have we fed jibri to voctomix yet? 18:18:39 #agreed we will be using Voctoweb and not OBS 18:18:40 CarlFK: I expect that will be caught during the human review in Sreview 18:18:43 one sec 18:18:45 tumbleweed: me too. some people say "oh no, we can't ask for more!!!" and that makes me sad :p 18:18:51 olasd: no, we haven't got an all-up setup in voctomix yet 18:18:55 olasd: no, I ran out of time this week to get Jibri setup 18:19:04 ack (just making sure) 18:19:05 oh 18:19:08 #action paddatrapper set up Jibri instances and ansible it 18:19:12 I'm not too worried about it but... 18:19:13 what's going to be used for the loop that shows schedule / time / potentially small videos / etc? 18:19:16 we might want to see about that first before utterly ruling out OBS then 18:19:36 Yeah 18:19:37 I thought that OBS might still be good for that (but perhaps a suitable replacement has been found for that already too) 18:19:39 highvoltage: ivodd had an idea to get tha tinto the sponsor loop 18:19:44 pollo: I guess we have to figure out when and how much review of the pre recorded video will happen 18:19:49 but unless this gets implemented... Maybe a VNC grabber? 18:19:55 tumbleweed: will that sponsor loop be able to show the time? 18:20:10 I think we could have a preview-equal with the Grabber and the llop 18:20:21 highvoltage: anything can be made to do anything 18:20:32 OBS is a reasonable fallback option there 18:20:40 true, although OBS can already do everything that we want 18:20:52 and there might not be enough time to implement all the features elsewhere in the stack 18:21:08 sure, but doing it wit hOBS means implementing glue 18:21:08 I just want to note that I think it's a bit early to throw away OBS entirely already 18:21:19 yeah, I wouldn't want to throw away the OBS VM 18:21:26 it's a useful thing to have on hand 18:21:26 for rendering text to video stream, gst have a few options that seem appropriate 18:21:39 OBS can just project into jitsi, easy cheap glue 18:21:48 projcet into jitsi? 18:22:07 how are you doing that? loopback v4l? 18:22:15 #agreed we won't rule out OBS fallback and some things that are not yet implemented in our vocto stack (e.g. sponsor loop with the current time) 18:22:18 you enable the OBS projector (right-click on preview window, open a projector window) and capture that by a jitsi instance in a browser 18:22:40 ah, I see 18:22:44 is that what you were using in mdco 18:22:46 I once tried that with v4l and it turns out using the obs project is both easier and works really well 18:22:57 nah for mdco I did something more convoluted 18:23:20 yay for overly complicated duct tape hacks 18:23:25 or something 18:23:28 I ran a VM on a laptop in the garage and just exported it's entire screen and connected via vnc when I wanted to change something 18:23:41 lol 18:23:54 highvoltage: right, that's more like what I expected 18:24:01 and I assumed we'd be doing things like that with VMs and VNC 18:24:07 #save 18:24:33 only reason I did it like that was because it was never a planned feature but people started asking for stuff so that was an easy way to get it going live 18:25:11 we can add a ! clockoverlay ! to the ingest pipeline for the sponsor loop 18:25:15 and we'll have time on it 18:25:17 tumbleweed: we could still do things with VMs and VNC for OBS / live loop, but we wouldn't need VNC anywhere between something like OBS and jitsi 18:26:26 olasd: that's great to know, what about other small things, like changing schedule, would we need to rebuilt the entire sponsor loop? is adding our backround music easy? 18:26:26 I'm not saying we shouldn't do OBS, but I do think sticking to the stack we know well and have worked with for a long time has advantages, too 18:26:32 highvoltage: EPARSE 18:26:52 (I'm sure all these things can be done and might even be easy, but I don't know how to yet and just saying we have a working solution) 18:27:18 i.e., if things break down halfway through debconf, the fact that we know vocto quite well can help us fix things, whereas with OBS we might need to look at things a bit more to get things up and running again 18:27:54 I think " 14:23:41 #agreed we won't rule out OBS fallback and some things that are not yet implemented in our vocto stack (e.g. sponsor loop with the current time)" is a good compromise for now 18:28:03 and we can move to something else :) 18:28:07 sure 18:28:16 wouter: fwiw, and not sure if you're following that, it's not OBS vs vocto, the part that I'm suggesting for OBS is a very specific role that will eventually go through vocto 18:28:35 pollo: *nod* 18:28:41 yeah, no need to drag this on further 18:28:47 #topic Streaming setup 18:29:29 if we have both an HLS and an RTMP stream, we'll need to make clear what entails what and help people choose 18:29:38 the docs need to be updated for that, yes 18:29:52 is it absolutely impossible to watch an RTMP stream in a browser? 18:30:18 if that can be done, that would solve some issues too I think, no? 18:30:21 there must be a way to feed it into a webrtc style low latency stream 18:30:41 but I don't know how those are typically done 18:30:42 I think it's around here on the list of priorities though 18:30:43 I was under the impression that WebRTC uses RTMP, too, but I could be mistaken 18:30:47 * olasd points at the floor 18:30:59 so many ideas down there 18:31:00 not natively, you need a js library to play them, not sure if free ones exist 18:31:23 presumably jitsi is using such a thing 18:31:49 anyway, it seems like we've made a step improvement in stream latency 18:31:59 and we're not about to reinvent the whole stack 18:32:14 seems like https://opensource.com/article/20/2/video-streaming-tools can work, it seems to suggest some changes in the nginx setup to be compatible 18:32:20 olasd: well, we can always keep adding more things, but I think it makes sense for us to look at ways of reducing complexity for the audience 18:32:29 otherwise we confuse people and that will detract from the experience 18:32:38 highvoltage: that's basically what we do 18:33:07 tumbleweed: I knew we were using nginx rtmp, I'm not familiar enough with it to know that we used the exact configuration listed on that page 18:33:11 I don't mind having more complexity elsewhere in the stack if that makes things easier for us, but *if it can be avoided*, I think it's nice to look at cutting out HLS streams as opposed to RTMP ones 18:33:42 highvoltage: that example is playing HLS 18:33:50 that is, cut out the HLS one if we want people to use the RTMP one, yada yada 18:34:26 if we had to just have one, it should probably be HLS rather than RTMP, given that the HLS one is far more widely useable 18:34:26 having said that, I'm not doing it, so if you say "it's too much work", then it's too much work ;-) 18:35:35 shall we move on? 18:35:36 tumbleweed: yeah, exactly (hls > rtmp) 18:35:55 you can #action me to update the docs for the rtmp option 18:36:08 #action olasd to update the docs for RTMP streaming option 18:36:13 (looks like the chair is busy doing something else :P) 18:36:29 just while you're working on streaming :) 18:36:33 #topic Advice/training for directors (/talkmeisters) 18:36:56 How much of our stack is stable enough to document? 18:37:08 heh 18:37:10 good question 18:37:19 I'd say start documenting, but don't take screenshots of all the things 18:37:22 Screenshots and such will change I'd guess 18:37:29 yeah, ok 18:37:39 +1 for writing prose and waiting before taking screenshots 18:37:42 documenting the architecture would make sense 18:37:52 and we can refine the details as we go 18:37:57 maybe stick to the general "try to prefer this type of layout" kind of advice? 18:38:00 it also helps locking some parts down and have a place to suggest changes 18:38:13 (for now, that is) 18:38:16 #action paddatrapper to start working on writing text documentation of our tools and our overall architecture 18:38:32 voctoweb is only a couple of days old, and we know we have big plans for things to change there 18:39:11 the Jitsi proceedures and setup is fairly stable 18:39:24 cool 18:39:25 And our pipeline is also 18:39:46 are directors going to have to interact with those? 18:39:58 wouter: those being? 18:40:03 I guess now that you've got jitsi up, I must bring up asterisk... 18:40:07 jitsi, pipeline 18:40:21 wouter: jitsi - yes, pipeline - hopefully not 18:40:25 jitsi would be used in BoFs and Q&A 18:40:37 right 18:40:39 so I'd expect talkmeisters to be driving it 18:40:47 +1 18:40:55 but, given they often don't exist. That could well be our directors 18:40:58 which means we need to document it -- okay 18:41:02 both would be a lot for directors 18:41:12 Actually two different docs may be helpful - a 'room layout' one for architecture and then director/talk meister doc 18:41:21 +1 18:41:48 #agreed split the docs into a 'room layout' one for architecture and a director/talk meister one 18:42:08 #topic Advice/training for presenters 18:42:09 "No recorded Q&A if there's no talkmeister... and we mean it this time" 18:42:17 #undo 18:42:17 Removing item from minutes: 18:42:21 lol @ recorded Q&A 18:42:32 heh that could well work 18:42:41 * tumbleweed knows what you mean, though 18:42:50 (anyway, that was just a quip) 18:42:52 #topic Advice/training for presenters 18:43:13 The docs here are done, and I think the link has been shared in the -team meeting 18:43:22 So, that's been published. Do we know if anyone has been using it yet? 18:43:29 I guess from #debconf-team, at least 1 person has recorded a talk 18:43:31 have talks been accepted yet? 18:43:45 no 18:43:55 (at least, not to my knowledge) 18:44:01 #link https://debconf20.debconf.org/talks/statistics/ 18:45:09 Anything else here? 18:45:30 I will have one more MR fixing some of the wording on the page, but nothing structural 18:47:09 I guess, judging from the conversation earlier 18:47:20 we'll need some sort of metrics for reviewing these prerecorded talks 18:47:58 and a team to do that. or will the speaker assistance desk be doing technical review? 18:48:39 can certainly help there, though I think it is going to need more people 18:48:59 tumbleweed: what kind of metrics are you talking about? 18:49:15 maybe technical requirements is a better word 18:49:20 do you mean like resolution and framerate etc? 18:49:26 right 18:49:36 i.e. if we're going to go back to the speaker and say, your volume is too low, we should say what we're expecting 18:49:40 and probably have a checklist for the team 18:49:59 yeah, makes sense 18:50:26 Shall we put a pad together somewhere to iron that out? 18:51:10 sure 18:51:19 something on storm.d.n? 18:51:23 yeah 18:51:32 After the meeting then 18:51:42 +1 18:52:01 though it shouldn't include audio stuff if I'm going to normalize that in SReview 18:52:25 #agreed we will create a pad with a checklist for people reviewing pre-recorded talks https://storm.debian.net/shared/qrG2PUSaMbSK2YK7JtdruXILp6VodsAcuQpbVFIs-Sh 18:52:37 then again, if audio levels are way too low, then normalizing it would probably result in loads of noise 18:52:51 wouter: well we still want the source to be recorded well enough that normalization has something to work with 18:52:53 right 18:53:15 right, which is why I'm bringing this up again now 18:53:27 the ideal is for the speaker to fix it in the recording 18:53:58 #topic Any other business 18:54:52 nothing from me... 18:55:05 nor me 18:55:09 neither I 18:55:24 #topic Upcoming meetings 18:55:35 IRC - next week Thursday at 18:00 UTC? 18:55:40 wfm 18:55:46 Wfm 18:55:49 sure 18:55:55 +1 18:55:58 *nod* 18:55:58 #agreed Next meeting: 23 July 2020 @ 18:00 UTC 18:56:05 Ah, before I forget: please have a look at https://veronneau.org/drafts/debconf-videoteam-sprint-report-debconf20home.html if you haven't already 18:56:17 Planning on publishing it later today 18:56:20 pollo: so one more thing, from before the meeting started 18:56:39 you seemed to imply that people would take part in BoFs while watching the live streams 18:56:46 surely they'd be in jitsi? 18:56:59 * highvoltage hops to the next meeting and waves goodbye here \o 18:57:02 might have something… 18:57:24 there is still a minidebconf planned in September ;-) 18:57:29 tumbleweed: hmm, I was referring to our on-premise infra 18:57:39 pollo: ah, :) 18:57:42 tobi: ok, we'll go back to AoB in a bit then :) 18:57:50 Test runs were looked at for this weekend. I suggest we push that out a week as we currently don't have a full pipeline running? 18:58:06 paddatrapper: yeah 18:58:06 AoB? 18:58:08 I don't have the spoons for it this weekend 18:58:17 tobi: Any other business 18:58:39 #agreed Postpone test runs a week to 24-26 July 18:58:51 #topic Any other business 18:59:18 tobi: so... your minidebconf 18:59:38 minidebconf regensburg … we still do not know if we can realize it,… 18:59:48 my only reaction to that is: "really?" 19:00:25 #info MiniDebConf Regensburg may be happening in September 19:00:38 so, in a few days we will open registration to get data if people show enough interest. 19:01:10 there are other boundary conditions as well, of course, but without data is hard to plan. 19:01:17 #info registration will determine if there is enough interest, opening in a few days 19:01:37 ok we just need to keep an eye on it then 19:01:42 tobi: if you do it, what are you looking for from the video team? 19:01:58 Do we have hw for that miniconf? 19:02:02 we don't have audio equipment any more :P 19:02:15 yes, equipment it is… 19:02:30 * tobi remembers faintly about the incident… 19:02:37 Hasn't it be replaced? 19:02:44 it has not 19:03:06 Slightly less urgent than it was, but we probably should look at that 19:03:13 * wouter doesn't know about an incident... 19:03:42 what would we need. Mixer? Mics? 19:04:41 Ideally one set of the audio in this section I'd reckon 19:04:42 https://wiki.debian.org/Teams/DebConf/Video/VideoTeamDocs/NewVideoTeamHardware#Kit_list_per_talk_room 19:05:13 (we still have the mixing desk) 19:05:53 wouter: the van with all the gear in it was broken into during the FOSDEM sprint 19:05:55 this reminds me I need to chase the opsis donation that DHL failed to deliver to my closed office during covid-19 19:06:18 Are the plans to replace them before September or should I check for rental opportunities? (The venue has some equipmen, but it IIRC is kind of hard wired with the room) 19:06:19 oh dear 19:06:31 tobi: I think you'd want to look at rental opportunities 19:06:45 ok. 19:07:14 and probably get cameras, opsis, and the PC from us? 19:07:24 well, I don't know; the audio hardware does need replacing and there's potentially more events in the fall (there's another MDC in planning in Bordeaux) 19:08:03 so, the replacements could be bought and shipped to the first one of these 19:08:15 if the shipping situation is sane 19:08:23 olasd: can you even get at equipment at this point? 19:08:27 if Thomann has those… 19:08:31 hmm, so thank you bxl :/ 19:08:36 tumbleweed: yeah 19:08:44 my office is open 19:09:04 AC is on so you need to wear a mask at all times, but I can extract equipment.... 19:09:09 but you're unlikely to take an equipment road trip? :) 19:09:58 Thomann is close enough to drive to for me, if that helps. 19:11:17 tumbleweed: right. 19:11:25 (though domestic shipping is probably cheaper than driving there) 19:12:17 I can look at making a thomann basket for the missing equipment 19:12:40 cool 19:12:42 #action olasd to look at making a thomann basket for the missing equipment 19:13:07 Anything else? 19:14:30 not from me 19:14:48 * tobi brought up his concerns… 19:15:29 tobi: thanks! 19:15:37 * tobi hopes that situations will permit to see you soon again IRL... 19:15:38 #endmeeting