# Opencast Summit 2022 – Session Notes ## Day 01 ### Opencast 11 * In earlier releases there was (afaik) no possibility to filter/search extended metadata via External API, is that still the case? * Yes, it is still the case * see PR 3274 on GitHub ### educast.nrw * would you recommend waiting with installing Opencast until educast is running smoothly? ### Roll Forward * have you thought of duplicating the published material too, instead of reprocessing it? * We do not archive publications. * @Manchester: When duplicating videos from year to year - are these full copies, so that this also duplicates the used space for the videos on the hard drives? * Doesn't create a problem because videos are deleted after 5 years. * Is the Upload tool part of OC or a UoM tool? * it's a UoM tool * Our site archives the published material, could we use the updated workflow to republish the processed ones? * Yes ### Challenges of supporting a lecture capture service * are you monitoring/do quality checks for the content uploaded from the users? We have a similar upload portal and noticed that sometimes users uploaded hole movies (particularly from theater studies). Any licencing issues? * lectures recorded "only" for dass-students: do the other lecturers/ students actually know, that lectures are being recorded? (since some dass-students might want to keep privacy about their needs...) * Do the "DASS students" subscribe for "podcast" once the term to all lectures or do they have to announce their needs for every lecture seperately, how do they get the media? By a portal / by mail / link? Anyhow a great service! * are most lectures @UoM recorded within Zoom are being uploaded to opencast later? * a recording includes audio and the presentation, i suppose. Does ist also include a video of the presenter? * What software is your video portal running under? * Maybe take a look at the Tobira session later * Maybe you could think about turning to the "self-service"-approach in your lecture halls with hardware that functions "as a webcam" in the lecture halls, like f.e. Vaddio 2x1 / Vaddio OneLink, so lecturers record and upload their own sessions. Saves a lot of management and hardware problems. * Just wanted to send you guys a high five! Keep up the great work - your lecturers should be very happy to have you! ### Vosk – Free Speech-to-Text * What is the accuracy of the subtitles compare to google and others? * Example (en; live): https://data.lkiesow.io/opencast/vosk-test-en.mp4 * Example (de; live): https://data.lkiesow.io/opencast/vosk-test-de.mp4 *  Is it possible to mix multiple languages? As an example German with English? * could you explain a bit more about the vtt+? what is it, which players can use it, etc... * does it cope with background noise ok? * Are these Kaldi nnet3 models? Do you train your own or are these copied together from other projects? * These are Kaldi models. We use the ones provided by Vosk * Does the data stay on campus? * yes, the software runs locally ### Subtitle2Go * https://github.com/uhh-lt/subtitle2go * Will you submit an Opencast module for integrating with Subtitle2Go? * No * community support is required * Can anybody edit the subtitles in the editor? If yes, do you have some process in place to prevent missuse? If yes, do you have some process in place if several people suggest different subtitle improvements? * at the moment only the uploader can edit the subtitles * Can you do this on live content? E.g. live-stream to Subtitle2go and get instant feedback? * there is a plugin for BBB without punctuation * bbb-live-subtitles: * https://github.com/uhh-lt/bbb-live-subtitles * Is it in any way possible to integrate this in Vosk * It should be possible to use Subtitle2go's nnet3 ASR model in Vosk. But our punctuation/segmentation are custom and cannot be directly transfered to Vosk (would need some glue) * Can subtitle2go be used through execute-once WOH? * may need glue script * Is editing subtitles by the uploader used very often? Do people actually take the time to improve them? * no data, yet ### Opencast Subtitle Editor * Accessibility and Keyboard shortcuts? Standards? * on the road map * What's required for the further development of such an editor? * currently work for and funding from educast.nrw * Is there a release date? * OC12 * Will the captions also been cut, if i cut the video? * Yes * Will there be a support for translation of captions? * as long as the language is configured in the editor, subtitles in that language can be created * Can you upload subtitle files from external sources (professional services)? * yes * Is the standard VTT? * yes ### Amberscript * It seems you broke your own integration in Opencast by updating your API but not updating the Opencast module you provided. Are there plans for fixing the integration? https://vt.uos.de/6ylyw * It's actually still working, just that the jobtype 'transcriptions' doesn't deliver the optimal output for captions. * However, we'll update the transcriptiontype soon! * How scalable is this?` * Very scalable, the ASR is running on servers with auto-scaling * For the 'human' service, there is a growing community of transcribers currently delivering hundreds of hours of captions per week - the community grows according to demand. * how expensive is the service? what price models do you have for academic institutions * It really depends on the volume you plan to transcribe. For higher volumes (10.000h+), the costs are below Google ASR * Just let carolina@amberscript.com know if you'd like to receive a cost estimate! ### Way With Words * You are very good with English/South Africa. Do you support German and other Languages? * We do for small-scale projects, however have not supported German at scale as of yet. * Do you plan on integrating with Opencast? * I believe we are integrated under the company name Nibity. Please contact graham.morrissey@waywithwords.net for more information on hybrid captioning (human/machine), single-speaker or academic model automation. For Saigen, please contact info@saigen.co.za. ### Ownership of Videos * These questions are kind of a barrier for us (HHU-Düsseldorf), which stops us from hooking onto educast.nrw, since in our mediathek we have also videos produced by the medical departement, containing patients data, which can´t be opened as OER.... * how is ownership of video an enforcing factor to open videos as OER? * [regarding educast.nrw from Matthias]: a tenant for an institution is _not_ automatically OER. The institution can set access rules. All OER content for ORCA is in a separate tenant. * ORCID provides a persistent digital identifier (an ORCID iD) that you own and control… * https://orcid.org/ * could ownership be transfered to other persons? ## Day 02 ### Virtual Capture Agents - if I understand correctly you already have some rooms and you additionaly created a "shadow" ones? - All room are “virtual” in the sense that there are no “real” capture agents - How does the matching algorithm works on created timestamps some minutes before schedule? - It matches as long as the times overlap. You can tweak it by providing a minimum match. - is it possible to create something like this but with scheduling? For example I like to schedule a recording for 2 hours but don't start the capture agent and assign any adhoc recordings started in the given room/time to a given series? - A: Right now, htese would get matched to the same recording. But you couild extend this functionality to just create a second recording instead. - A2: actually, if you provide a file to the OC-Inbox with the expected metadata format according to the metadata-extraction config, it should work out of the box - I don't like to use the inbox :P. I want to schedule and then ingest...if it is possible :). Starting an adhoc recording will start a new workflow with only the location info available - I see :) Actually I was also thinkig about that (eg also for missed scheduled recs, to be able to add them to a scheduled event manually) - Sorry, missed the beginning of this talk - do you use this virtual capture agent also for lectures out of a lecture room? How do the signals (presenters Powerpoint and Audio) are plugged into your TU-Graz-Ethernet? - A: we use Epiphan recorders that send the recorded files directly to the OC-Inbox - what happens if you start and stop in the scheduled time --> you have eg.: 3 events that match? - A: Problems :) – but managable; we're still assessing those cases, I will be able to report more at a later point - what happend if the adhoc never starts? Is it staying at "scheduled" and you delete them manually? - A: Yes ### pyCA - https://opencast.org/2022/03/08/pyca-4-5-has-been-released/ - livestream feature? - A: it is possible - Is there a roadmap? / What is planned? - A: in Lars' brain-repo only but feel free to fill an issue at https://github.com/opencast/pyCA/ ### Adena Arec - Do you also have speaker-tracking cameras, without external microphone systems? - Answer: Our cameras do not track by voice, unfortunately. But the IR solution comes close to what you are looking for :) - Unfortunately not, as we want multiple speakers (10+) in a conference room to participate in a VC without somebody operating the camera - I see. In this case I'd recommend to get one of the integrated microphone systems. We don't have cameras that would be able to do that on their own. - Does the tracking also work with ceiling mics? (e.g sennheiser tcc2) - Answer: Ceiling system from sennheiser is not integrated yet. - Are all recording devices compatible with Opencast? - Answer: Except LS-2, yes. You can see the full list of our devices on www.a-dena.com - could you name some prices (media station) for the german market? - Answer: please email me at boris@a-dena.com and I will be able to support you for that. - Your LS220 supports 4K? - Answer: I believe you mean LS-110. It supports 4K for input and local output. Recording and streaming is done in Full HD. - Is the small USB and the touchscreen device compatible with Opencast? - Answer: Yes - Please feel free to address any further questions to my email address boris@a-dena.com. We are able to arrange a product demonstration for you and support in other ways! ### Extron - 4K coming? - A: Definetively on our radar, can't commit on a date due to the current component shortage crisis. - Opencast eventID and or seriesID in livestream-URL? - A: On our radar for upcoming FlexOS app releases - Livestreaming only if activated in Opencast and not for every recording? - A: Currently not possible - Will submit this as feature request. - Will it be possible to control the SMP layout presets via Opencast? +2 - A: It is possible using iCal through a custom key. We should explore together if this is doable using Opencast feature set - Official release of the Opencast App? - A: Released and available on the website - Does TLS 1.2 and 1.3 now work? - A: Will check the status and keep you posted. - Is ansible support planned to manage the units? - A: Currently not possible - Will submit this as feature request. - are the virtual inputs limited to FullHD? - A: Yes - 4K coming WHEN? - A: On Smp 35x I do not expect Virtual inputs to be 4K compliant. - About the the 4k capable hardware, this is on our radar but we can't commit on a date due to the current component shortage. - Thanks all and for any other questions feel free to reach out : abenni@extron.com ### Capture Agent Operations - A lot of points where about monitoring devices. Would having Prometheus metrics for video interfaces help? In additiona to metrics about the capture agent. - Test recordings to check for a/v sources - https://github.com/opencast/pyCA/issues/382 - what about Linux-Hardware? did you test CaptureAgents that are "Linux boxes"? with which video card? - all are mini PC's running Linux - acquire too many cards due to break/fix support method - Blackmagic - cheap, no scalar-fuzzy about input signal (extra scalers needed) - Datapath - expensive, support not as good as it was - Epiphan USB - USB is flaky - Magewell - best soln going forward, but custom drivers- magewell.com/products/pro-capture/list - UPV devices shouldn't have driver problems. Did you see any problems? - Epiphan AV.io, … - Other USB UPV capture cards - do you have "no go" or even red lines in decision making? Budget ? Open Source software? Why not just go for an appliance? - mostly for flexibility and custom stuff you couldn't easily do with a commercial appliance, but obviously budget can be an influence when you start scaling up! - External H/W eg scalars, HDCP management ;-) - Try and get AV people to turn off HDCP on at least the laptop HDMI input - Use a capture card with built in scaling if possible - @Paul + UoM - Team: Thank you a lot! ;) ### Easier hybrid teaching - Opencast Capture Agent - Extron and Arec Media Stations - What about (live-stream) video quality in case of blackboard use? - it is used now, but we have to see how good HR cams working - I'm interested in using WebRTC in general as ingest/delivery mechanism for streaming outside of BBB i.e. "normal" (embeddable) players. Will you work on that? - hint to Carlos' presentation tomorrow (12:00 UTC) - WebRTC is a "wild west" as each product is implementing a custom flow. Do you see standards emerging? What about ingest via FFmpeg / OBS etc.? - WebRTC seems to be less problematic than HLS on our side - do you have ideas for the interaction between "virtual" and live students - Chat, Shared Notes, votings. In general our Videoconferencing system ist openen for all students too. - or interaction between students and lecturerer - schnaq is a new Q&A-tool by students of HHU: (startup) https://schnaq.com/de/ - It does not seem to be free and open source software? - Teacher should have option to see, their remote audience. As the balance between remote and local is unknown. - Answer: Is think so too, there should be a second screen in prepared rooms to show them - university - bandwidth: HHU Düsseldorf just moved from Vodafone to the dfn-network now we have 2 x 5gbit/s... - We only have 3GB at UOS - we (HHU Düsseldorf) have a license with cisco webex, installed cisco roomkits in 6 lecture halls, have "support" from an external company, but: the support for teachers is useless, we still have bugs in the roomkits, the installation was a pain.i.the.A...(cisco ptz-cams needs 3 cables-HDMI, RJ45+current for instance)-Means yes - go for it, use BBB and opencast - I have over the years my experience with videoconference defices from Polycom, Lifesize and Sony. The support was never very good, you have mandatory quite high maintainance subscriptions. ### Opencast Zoom Integration - Does a Teams integration could work the same way? - MS Teams currently has no API for extracting recordings. It's still in beta. - Can be the Zoom transcription module used for the recordings and how would it be ingested to Opencast? - Aren't zoom transcriptions done live? therefore lesser quality. A: Yes, that might be the case, though 1:1 time seems to be the standard anyway. Using more time doesn't increase quality necessarily. ### Beyond Elasticsearch: User-facing search engines - Do you want to replace the Elasticsearch in Opencast with this? - Features are greate, but what about prod. environments? HA? Sharding? Does it really scale (out)? - How secure is the future of those projects? - Is Elasticsearch really slower or just in theory? - Yes, using fuzzy search in prod video portal (UoM) - Community support (for deployment)? ### Opencast Architecture and Dependencies - Was this helpful? If so, let us know so we can do this again next year - Yes it was ### High Availibility - Making video delivery always work requires also that metadata is available via the (external) APIs for LMS - We need to think about every step in the pipeline. CAs, Moodle and I think ILIAS already provide a cache and can just retry uploads later on. - Workflows/jobs don't always recover correctly from a restart. - Resilience if a service (DB) is temporarily unavailable - Memory usage during rebuilds - How long would rebuilding an index at a "larger" institution typically take in absolute figures? ## Day 03 ### Paella Player 7 - Where is the new repository? - Core Library: https://github.com/polimediaupv/paella-core - Player for Opencast: https://github.com/polimediaupv/paella-opencast - Pull request for getting Paella Player 7 into Opencast https://github.com/opencast/opencast/pull/3162 - Will the current configuration files for PP6 be compatible with Paella Player 7? - Yes - Will there be a backport of Paella Player 7 for Opencast 11? - Yes, no problems - Is Low Latency HLS supported in Paella 7.x? What Low Latency HLS “standard” (Apple vs. Community)? - Yes, there is even a demo in the paella player website. We use CMAF - Will 360 degree be supported? - Yes, but not in short term. I'll work with Ypatios for that - Will an end-user configurable quality cap be available for HLS (for metered connections)? - There is a configurable quality setting, maybe that's enough - Does Paella support HLS audio groups (HLS "Alternate Media"), i.e. late binding of video and audio tracks? - Does Paella support subtitle tracks in HLS manifests? - yes - Will you improve HLS ? Because a lot of students with lower bandwith is not able to watch HLS through Paella7, but using the same from video-js or plyr player - We are working in that - What about timeline previews (vs chapter previews)? - Not in the short term, but it is under the scope - If both videos have audio, which audio is used by default? - You may configure that. However, that is not possible for iOS, which means having two audios is not a safe practice - If both videos have audio and the used one is closed, will the closed one be unloaded and will the remaining one play it's own audio? (Thinking on syncronization problems and slow connections) - I would said yes - Where can I find the Wowza config files for HLS-LL and the corresponding config file for Paella Player 7? - In the Paella player website in a few days. I will send also a message to the OC-Users list when ready ### Tobira - How to develop integration into authn/authz systems? - You can add your own plugins for that - yes but how? Can we start working on that now? - In theory, yes. But iot's still alpha. This means that things could still change. Even if that is unlikely for that part. - The system is very similar to what Grafana does - What dependencies does Tobira has? - PostgreSQL - Later probably search index - Will you work on High Availability? - Yes - Is there already some documentation (technical, admin, user)? - Only what is in the repository. But there you will already find some. - Is Tobira still beta? When will be the first release? - No. It's alpha :D - Extend upload with multiple files (videos, subtitles, preview images)? How extendable is this? - Not now. On the roadmap. No deadline for that yet. - Subtitles can be uploaded in the video editor/subtitle editor. - How extendable are the metadata fields? A: Remember ETH is involved... - Is it possible via tobira to have users fully manage the rights to their videos and content? F.e. granting certain groups of people access to videos or allowing other members of staff to manage and/or edit their videos? (like in a scenario with a professor creating and uploading videos and a student-assistant editing the video) - Tobira adheres to Opencast's access control lists. So, you can have multiple people allowed to write - Additionally, remember the ownership discussion ;-) - User will, at some point, be able to edit ACLs in Tobira. But probably not like in the admin-UI, but more simplified. - Is the uploaded content available in OC's backend aka admin UI or is it a special "section" - All media related data are available in Opencast. - What is the current timeline for a first beta and stable release? - around August - Can we separate out the cache for e.g. LMS? -> extend HA to those systems as discussed yesterday. - It's a bit hard bc our database schema is quite interlinked. For example, series blocks on pages have "foreign key" contraints to the series table. So I think not. - Can we get rid of the stored procedures in the DB? - Everything is possible, but rather unlikely? We chose to use the DB as smart system, not as dumb DB. It would be quite a bit of work to now change that approach completely. - Can videos be added to multiple (play)lists and/or pages? - To multiple pages: yes. But currently there is no notion of "playlist" yet, so no. But certainly planned to have some n:m video list - How are ACL handled,ie login mapped to rights to view or manage videos? - So Tobira uses the ACLs of events and series. Tobira expects information about each user from the opaque "login handler". That login handler has to provide a list of roles the user has. Those are then checked against ACLs of events/series. - Q: So it is using the database to apply the ACLs? Does that scale well if the user has hundreds of ACLs (eg program admins) ? - A: yes. In the database, we store the ACLs for `read` and `write` as string array. We can then check that in SQL by using the `&&` array PostgrSQL operator which checks for overlap in two arrays. So far it works fine. PostgreSQL even offers an index for "array overlap" operations, which we use. - Links: - Repo: https://github.com/elan-ev/tobira - Docs: https://github.com/elan-ev/tobira/tree/master/docs ### Status of the new Admin UI - repository: https://github.com/TIK-NFL/opencast/tree/admin-ui-picard (the branch is important) - Is there a timeline for when the new interface will be released? - probably this year - Which Opencast version might picard be targetted to? Can it be adopted to older versions of Opencast? - I would not have expected the adapter registraion via the circle icon with the question mark, I would expect a help section with the question mark - It was like that in the old UI, so we kept it that way (it could be changed in the future if that is a wish of the community) - so no UI/UX activity from your side? (A: we are no UI/UX designers, we are developers, we try to get the same UI as in the past, but programmed using React.js instead) - ok I see, thanks! - we only did minor changes more concerned with functionality, like adding save buttons on the details modals, so that users don't make changes by accident, so far - Presenters are realized with a box around, other links in the admin UI not, so for me it is not consistent - had you a look at UI/UX also? Same for the icons and other UI elements. - As mention, above we not changing the UI, however some changes concerning consistency could be possible in the future - For now we are aiming to get a running version of the new UI before making beauty changes ### Plugin Management - Karaf Features: Every dependency management concept just creates new layer of dependency hell ;P - What about getting rid of distributions and make this just configuration? - This will come with OC 12? - Very likely yes ### Annotation Tool - Recording from OC-Summit in Vienna, 2018 (1:58: Live-Demo of basics; 14:31 Scenarios): https://ustream.univie.ac.at/media/core.html?format=hls&id=30d3c9e6-43ad-445d-9825-d00461cb6a27 - @ELAN when will there be a merged community version? - We've spent some time over the last two month to sift through non-community patches and see what makes sense to bring it in. This should be done in the next few days. We also made sure that the tool now works with Opencast 11 and fixed some additional bugs (Lars) ### LMS PHP Library - Is the ingest API included already? - I think only supported OC versions should be supported (i.e. the last two). Maybe also link lib version number to OC version number. - Development made for StudIP and Moodle. What about ILIAS? - Farbod:it is library to be used by all - Which PHP version is supported? (PHP 7.4 is security support only, 7.2 is EOL upstream) - Farbod: as of now 7.0 - If is in ongoing development why not 8.1? - Support for pagination/continuation? - Farbod: of course - Do you have a git repository? - Still empty but here: https://github.com/elan-ev/opencast-php-library ### Open Board Session - Next conferences - German speaking meeting in the summer June / July 2022 - Media and Learning conference https://conference.media-and-learning.eu/ - Opencast Community is member of the M&L association. Discounts for members might apply. - Additional online summit: We don't know yet. https://opencast.org/2021/11/08/new-opencast-summit-schedule/ - Become an Apereo member to support Opencast: https://www.apereo.org/ - Workshop - Where to meet: meet.opencast.video? - We will announce that - Could you put the old conference dates on the website -as our memories are failing! ### The Great Manchester AWS Migration - Do you authorize users with CloudFront? Tokens? What about users keeping the player open for days? - How do you package videos for delivery? HLS/DASH/..? - just pure mp4 (720p) - Q: aren't large files inefficient for CDN caching? Can CloudFrond cache byte ranges? - A: not sure about cloudfront caching ranges, but files are not too big and most people watch the whole video anyway - What about AWS and data privacy regulations (in the UK)? - We just have to use London (eu-west-2) region - Did users notice the migration, e.g. faster video start time, less buffering etc.? - Have not had time to analyse, need users who used teh service before and after. - Not had any compliants about playback speeds - How do you deal with mediapackages in archive already moved to S3 that then change e.g. setting metadata / starting a workflow? Especially when changing metadata the whole mediapackage needs to be downloadad and snapshoted again. - yes, it is just as any other storage, we pull current version down to EFS and push back to S3 when processing is finished - Q: is this not a problem for latency when a users changes multiple metadata fields? - A: actually, i might be a bit wrong there, maybe someone else has a better idea than me ;) though I think our PHP tools upload catalogs to the archive then run a republish workflow so yes that probably happens. we don't push to S3 when a mediapackage is awaiting editing for example (i think!) - Do you scale workers with the load? I remeber you talking about a fixed schedule when scaling workers. What metric are you using? - we scale both ways, james can elaborate - CPU - we change the autoscaling limits for evenings and weekend - videos can take longer due to less resources available. - Have you considered other CDNs? - We weren't expecting to pay for data egress! (shh!) - Also egress from S3 to Cloudfront is free - What's your current backup and recovery scenario? - how was the deleting policy before moving to aws? Did you have to move lots of material that is actually not being used any more...? - Yes, we moved everything, we did then perform our year removal of old lecture videos (>5yrs) - We have a lot of useless videos, but working out what these are is task we don't have time to do. Also if one student needs that video then we would still need to retain it. Storage costs vs human - What machine configuration do you use (esp. workers) (CPU, mem, AWS type)? - workers C5.large - admin runs on M5.xlarge - mainly to handle reindexing, we could reduce this, tho costs saving are minimal. ### Opencast Ansible-Roles - https://galaxy.ansible.com/elan - https://github.com/elan-ev/opencast-tobira/tree/main/.deployment ### Opencast in AWS (and other clouds) BoF - https://bitbucket.org/uom-media-technologies/podcast-terraform ## Workshop planning - Accessibility: https://terminplaner4.dfn.de/OgAMMmejgGOzQLYM - Friday March 18th, 9:00 UTC - 3 Participants - Workflow Configuration in JSON: https://terminplaner4.dfn.de/79Ywu2mNK4UcEyTX - Wednesday March 16th, 14:00 UTC - 2 Participants - Integration of Opencast in Moodle: https://terminplaner4.dfn.de/e5B1wchRC6QWHYsh - Thursday March 17th, 10:00 UTC - 14 participants - Integration of Opencast in ILIAS: https://terminplaner4.dfn.de/oi854qLjz5DFkwGN - Tuesday March 15th, 12:00 UTC - 13 participants - Opencast Annotation Tool - A common way forward: https://terminplaner4.dfn.de/sLizRIT3GxluAZ4R - Thursday March 24th, 8:00 UTC - 8 participants - Ownership of Videos - Date still hast to be found