¶ Talk Recording
Quick Start: Refer to Basic OBS, Simple Setup and Actually Using the Setup if you only have 5 mins to learn how the setup works.
Note: This guide assumes some proficiency in OBS, at least enough to switch scenes and check sources and (Source Record) filters.
¶ The Tech
Doing WASD has graced us with plenty of AV tech, but most of it is entirely unecessary to record a talk.
Unecessary Tech
- ATEM Mini Pro ISO: Lovely in theory, but it can't do the compositing we want, and requires straight HDMI in (no passthrough)
- Can record each source and generate a DaVinci Resolve project for the cuts, also livestreams out, so nearly what we want
- also has been used as a £400 HDMI extender
- XR18 Mixer: Lovely tech, but we don't roll with XLR mics enough for this to be worth it (we did use it as a glorified XLR converter once tho)
- controlled via x-air app on all devices over wifi (best to create a LAN without internet as internet messes with things)
- if you want a physical mixer surface ask Joel (Mole1424)
- 3x Behringer Handheld Mics: XLR so same as XR18, decent quality tho
- 2x LED Light Panels: If you want to properly light your speaker, sure, but also not worth the effort
- Many XLR/aux cables, converters adapters
The equipment we usually use:
- Elgato HD60X Capture Card: For recording the speaker's laptop
- Room PC for access to Lecture Capture Mics and Cam
- Rode Go 2 wireless mics (optional): use for speaker and/or questions
- The Camera (optional): AWOL half the time, but much better quality than lecture cam
- Owen's crappy non-passthrough capture card (optional): capturing camera input
¶ The OBS Setup
OBS is powerful and not too difficult to use. I recommend getting familar before you have to use it to record a talk. OBS is not installed on the PCs by default, but is easy to use portable. My portable version with the preset scenes is available at aaj2005.sites.uwcs.co.uk/OBS.zip For reasons known only to God and ITS any website hosted on warwick's internal network cannot be accessed by any of the windows computers in warwick, hence you will need to save the file on a USB. Furthermore, the uncompressed file is too big for a standard warwick network user area, to combat this it needs to be saved locally (the videos folder is what I use). Make sure the recordings get copied off (onto network/USB), and don't stay local, then you have to go back into the room to get them (and they might be wiped at some point).
¶ Basic OBS
- The most basic operation is not caring about the composited screen.
- Change the inputs sources for cam, screen and mic,
- ALL THE SOURCES ARE RECORDED SEPARATELY: CHECK SOURCE RECORD DESTINATION EACH TIME on both screen and cam
- This will record the main composite stream and the screen and the camera separately for future editing
- Audio is currently only attached to the main stream, refine this in future (each source is a separate audio channel)
¶ Explaining the Scenes
The actual capture devices for the screen/cam/2nd cam in the scenes have a Source Record filter on, which is a plugin that records a separate video file for that source. This is very useful when editing afterwards, unsurprisingly it does use for CPU though. If you set one up for yourself, set Record Mode to Recording (so it records when recording), and add a suffix to the file name to denote which stream it is.
Audio is separated (probably unnecessarily) into audio channels. These are all attached to the main video atm, but allows mixing the levels of them all. Channel 1 is a mix of all, then main speaker mic, then the screen capture, then secondary mic and finally camera.
-
Four actual video scenes to use:
Cam L: Composites camera (portrait) on left, with screen smaller in the middle. Uses overlay and title to fill blank space.Cam R: Same but mirrored, depends on which side the speaker is standingCam Only+Screen Only: Super self-explanatory. They have a 1x1px version of each other for the audio sources.- Source record filters are on the scenes, this should work fine wherever you are
2nd Cam Only: If you have two cameras going (e.g. lecture cam and actual camera), use this as a cut away to the 2nd), rarely used, but worth having.
-
There are three "component" scenes, that you should never switch to on recording, but are composited in to the actual layouts
_Screen + Audiohas the screen CC (change this source here) and screen audio. Hidden sources in there can record that PC instead_Cam + Micrecords the camera (lecture capture or actual camera).
- ChangeCam CCto the lecture capture cam or camera capture card.
- TheSpeaker Micdevice to whichever primary mic we're using (room/Rodes). If you want both (e.g. room for speaker, Rode for Qs, enable the hidden source)_Titlestores the talk title (used in various places, easier to configure in one)
- Either set the text fields directly, or save an image into theAssets/Titlesdir to access in the (frozen) slideshow (for quick changes)
-
One component scene, just for static assets:
Title Slide: Places the title scene over a generic title slide (good filler)- One day maybe we should make a NodeCG bundle for it.
¶ The Simple Setup
We use OBS on the room PC, both for access to the lecture capture gear, and more reliable (and probably more powerful) than someone's crappy laptop. The speaker uses their own laptop, for ease of changeover/prep and demos are frequent, which is captured by the Elgato CC. Check the room beforehand, but the lecture cam and mics are usually capturable on the PC. This makes setup super easy:
- Walk into room
- Plug in Elgato CC Out to laptop HDMI, USB into Room PC, In to speaker's laptop
- Login and lauch OBS on Room PC (from USB/Network Drive)
- Swap sources to ensure capturing cams
- Fullscreen project/switch room source (see later) and hit record
Rooms in this list (requires exec on uwcs discord to view) are the best for a recording as they are "MS Teams" compatible, this means that the room's PTZ camera and mics are directly wired to the PC: meaning they can be easily selected as input sources.
¶ Adding the Camera/Mics
Some rooms don't need amplification or have problems with mics (e.g. MB0.08 has loads of feedback if speaker is standing at the desk). So then it's best to use our mics (if they are charged). Some rooms don't have a capture cam, or it isn't connected. In these cases, we have our own.
To add the mics:
- turn on the reciever and an actual mic (hold down the Rode symbol button to turn on)
- Plug in to mic port on Room PC (or with splitter in the case if a headphone/mic hybrid)
- Recently, I've been experimenting with plugging this straight into the camera (comes through on camera CC) - Change mic input source
To add the camera:
- Get a tripod (hopefully the actual camera ones have been found by then)
- Mount camera to screw on top, point at speaker (keep wide angle if possible)
- Plug the long HDMI cable into the side HDMI on the camera, then into small CC and into PC's USB.
- Change camera source to CC
¶ The oh shit set up
Some rooms (like CS1.04) are really annoying, not only not being MS Teams compatible but also the only sensible places to mount cameras so are too long to run a HDMI cable easily. Therefore the best way I've found is to set up everything on your own laptop as follows:
- Set up camera on tripod and point at speaker
- Run HDMI out of cam and into HDMI in on elgato
- Plug elgato usb into laptop
- Double mic speaker with room mic and rhode mic (plug reciever into camera audio input (above hdmi port on camera))
- Ask speaker for a copy of their slides (if live demo ask them to screen record and send over after the talk)
- Switch screen input in OBS to your own device and check levels on audio
- Press record
- Full screen whatever presentation speaker is using and mirror them on your laptop
¶ Actually Using the Setup
The Simple Setup instructions work a lot of the time, but if you are marching into a room with 5 mins to setup, it's worth taking a couple more to check all the asset paths, record paths, and sources are correct -- getting it wrong is far more of a disaster than starting a touch late.
¶ Screen Mirroring
- Some rooms have a super annoying property where the projector is mirrored to the monitor.
- Ideally, the projector will be on the speaker's laptop as Laptop (HDMI) -- through the Elgato Capture Card.
- So the projector gets the speaker's screen, the recording gets the composite, and the monitor gets OBS (for you to control)
- However, if the monitor and projector are mirrored, you need to fullscreen project the Screen CC source to the monitor (so you can't control or check the recording)
- (Right-click the screen in your composited view, click
Fullscreen Projector (Source)near bottom) - You need to set recording before this, or tab out (Alt+Tab) and set it going
- I have a few ideas to get around this, but nothing great (see Getting Around the Mirror)
- (Right-click the screen in your composited view, click
¶ What to Tell Speakers
To make the most of this, the speakers have to keep a couple things in mind:
- We are cropping the video to portrait, so they can't pace too much (the source records the full width, but the composite feed is cropped).
- I tried a face track plugin before, but it wasn't very reliable, and only worked in good lighting and head on.
- If they are pointing to the screen, we need to cut to the wide
- Ask if they need the doc cam or whiteboard - these are a little more interesting and we need to be ready with spare scenes/cams if so
- Ask them to repeat questions into the mic, so everyone knows what has been asked (and more importantly, so does the recording)
- Make sure they know they are using their own laptop, and either you or them bring a USB-C adapter if required (don't trust rooms will have).
- Obviously make sure they are ok with being recorded and their full name attached - e.g. some people don't want their face or name
¶ How to edit a video
This is gonna be a rough guide because it varies based on the style of video you want but here goes:
I mainly use davinci resolve, it is free and powerful + there's loads of tutorials online.
In general, try and edit out long uncessary pauses, but don't edit out every erm as this can soon sound unnautral. A good way to mask cuts is to change to a different camera angle, for example switch entirely to presenter view or the slide being switched to.
In some cases (for example revision sessions) speed of edit is better than quality. For these I edit out periods when the audience is doing the question and then editing in whatever I think the best representation of the answer is (for example the speakers answer at the time, or just a screenshot of the markscheme).
Thumbnails and titles of video are standard YT stuff. Try and be catchy but remember the primary aim of our content is educational and a record of lightning talks for future refrence so don't go overly clickbaity.
FOr reasons unbeknownst to man kind people really liked a python talk video that we posted. This means we have >1k subscribers. So after a long hassle with YT and the SU (with a sprinkling of borderline tax fraud) we can now monetise videos. This comes in every now and again and can be checked here. This is sent to main SU central account so you will need to email finance to let them know so they can transfer the money to us.
¶ Streaming(?)
The easiest place to stream is Discord, just login on the PC (ideally to UWCS account or someone with Nitro), start OBS Vitual Camera, and stream that source.
To stream to YouTube/Twitch, you need a stream key from them and put it into OBS, many guides exist, though iirc YouTube once made us wait 24 hrs before it gave us a stream key. If you wanted to stream to multiple (this gets technical), setup an RTMP server (e.g. SRS or Nginx RTMP Server) to forward the RTMP to YouTube and Twitch, and stream to that instead.
IMO, streaming is rarely worth it. It just discourages attendance, whereas videos don't seem to that much, and are far higher quality. It is also far more of a faff, with far more to go wrong.
¶ Getting Around the Mirror
I've had a few ideas for this. There are plenty of OBS Remote/Web control sites/programs, however the real problem is getting network access to OBS.
I've tried various tunneling methods (e.g. Ngrok) with unreliable and variable results, I've thought about SSH reverse port forward tunneling to DCS/UWCS servers from the Room PC, then connect with a port forward tunnel to the server from my laptop -- haven't got this figured out yet tho.
But I've found relying on uni networking a bad idea, so I've found the most success connecting a small macro keyboard (could do a Stream Deck if money), connected to hotkeys on OBS (Ctrl + NUM keys). This works for control, but not viewing, so the best solution I have atm is stream to myself on a 2nd account.
This keyboard is laid out with each button corresponding to the NUM 1-9 keys, with actions in order of 1-9: Cam L, Cam Only, Cam R, Title, Screen, 2nd Cam, Record, Change Title, Transition. 1-6 are for scenes (Title & 2nd Cam are rarely used, but maybe). Change Title is for when you need to change to the next talk, but don't have physical access.
Note, I do this on Studio Mode, which gives a preview on the left, and live on the right, so if I hit the wrong key, I can undo before transitioning to live.
¶ Python Talk LIB2
Everything OBS-wise is set up before hand. Should receive tech bag from UWCS exec.
Plug USB stick in to PC. Plug USB end of capture card (CC) into PC. Plug room HDMI into HDMI-out of CC. Plug other (included) HDMI into laptop (there are adaptors on the desk)
Start OBS by going into /OBS/OBC-portable/obsportable.exe (path might be slightly different I'm doing this from memory on bus). Select scene as Cam-L. Using the control panel on the desk, move the camera so that subject is framed in camera view on left. Use room mics as microphone (check levels in OBS beforehand)
When ready hit "start recording", then "stop recording" on finish. Make sure to navigate to Videos on room PC and transfer the camera footage to the USB.
You're done 