OPLSS Videography

This document describes the equipment and procedures we use to record the Oregon Programming Languages Summer School sessions and publish the recordings.

A/V Personnel

The OPLSS coordinator should designate a lead videographer and an assistant. The lead is responsible for all aspects of operating and maintaining the A/V equipment, as well as editing and publishing the recordings. The assistant should be familiar with this document in order to record sessions if the lead is unavailable.

A/V Equipment Inventory

We used the following audio-visual equipment to record OPLSS sessions:

• Canon EOS 80D digital SLR camera kit -- body, lens, batteries, charger -- borrowed from the university's media center
• at least three SD cards, each at least 128 GB capacity (for recording and storing videos)
• Fifine wireless microphone (connected to the camera)
• Shure SLX-1 wireless microphone, borrowed from the Knight Law A/V department (connected to the room's PA system)
• Manfrotto Compact Advanced tripod, also borrowed from the media center
• headphones (to monitor the camera's audio input)
• alkaline AA and AAA batteries (for the microphone transmitters and receivers)
• power strip
• HDMI video adapters:
  • USB-C
  • Apple Thunderbolt
  • Apple Mini-DisplayPort
• laser pointer

The videographer should supply computing hardware for post-OPLSS video processing.

Device battery inventory

The equipment uses a variety of batteries. The videographer should maintain the spare supply in a way that anticipates failure and rapid battery replacement.

Battery inventory:

Device	Battery Type
EOS 80D	rechargeable
Fifine mic transmitter	AA
Fifine mic receiver	AAA
Shure mic transmitter	AA
Shure mic receiver	AAA

The mic transmitters come equipped with indicators of the remaining charge. They typically run at least three days before falling below 50%, at which time they should be changed.

The Fifine mic's receiver does not indicate battery charge and will fail without warning. It typically runs at least two days.

OPLSS venue

OPLSS used room 175 in the Knight Law building, an auditorium that seats about one hundred. The room was equipped with a projector and public address system. A locked side room contained audio-visual equipment and served as storage for our valuables. A touch-screen panel adjacent to the equipment room controlled the room's projector and PA system.

The room's acoustics absorbed the instructor's voice, especially when facing away from the audience. We therefore used a wireless microphone connected to the PA system. This entailed asking the instructors to wear two microphones and transmitters. Most complied, but several refused.

The law school has dedicated A/V and IT staff who can assist with issues.

Pre-OPLSS rehearsal

At least one week prior to the first OPLSS session, the videographer should arrange a time with the event coordinator to visit the room and rehearse recording a session. This is vital to understanding the room's layout, lighting, and A/V control systems.

The coordinator invite the Knight Law A/V lead to the rehearsal. The coordinator should bring the A/V equipment (camera, microphone, and tripod).

The videographer lead and assistant should do all of the following during the rehearsal session:

• Mount the camera on its tripod
• Connect the wireless microphone and headphones to the camera
• Record someone speaking with the wireless microphone
• Practice unplugging the camera microphone, recording with the camera's built-in mic, replacing the battery, and reconnecting the mic. This is necessary because the camera's microphone will occasionally fail due to battery exhaustion, accidental power-off, interference, and other unexpected circumstances.
• Understand how to use the PA system mic
• Monitor the recording with headphones
• Cycle the room's projector controls. Connect a laptop to the projector.

Preparing to record a session

Arrive at the lecture room at least 15 minutes prior to each session's start time. Do all of the following during set-up:

• Prepare the camera:
  • Mount the camera to its tripod
  • Power it on
  • Verify the card's storage capacity
• Insert a fully-charged battery and charge the depleted battery
• Set the zoom and focus
• Connect the wireless microphone and headphones to the camera. Power on the transmitter and check its battery level. If below 50%, replace the batteries.
• Power on the mic transmitters and the room's A/V system
• Verify the camera mic's connection to its receiver -- when successful, the receiver's status light will glow solid orange.
• Verify that the PA mic is broadcasting through the room's PA system. If the speaker plans to project and needs a laser pointer, place it on the podium.
• Assist the instructor with fitting the lapel mics
• Stage a pair of AAA batteries by the camera in case the receiver fails mid-session
• Stage a fully-charged camera battery by the camera
• Using the headphones, verify that the instructor is audible via the camera's mic

Recording a session

• Start recording when the instructor begins speaking, ideally slightly before
• Record events that occur during the session:
  • The length of each video file. See below for notes about the camera's 30-minute video length limit.
  • Camera or microphone problems
  • Any other events relevant to post-OPLSS video processing

Framing, pan, and zoom guidelines

The videographer should operate the camera in a way that maximizes the instructor, white board writing, and projected content in the frame, while also minimizing camera motion. This as an art, as instructors vary in how often and how far they move laterally. Instructors who move frequently and unpredictably present the greatest challenge, sometimes requiring rapid panning. To the greatest extent possible, pan and zoom strategically to anticipate the instructor's lateral moves.

When the instructor is writing on a board, the operator should include the instructor and the relevant board content in the frame at all times. The zoom level depends on two primary factors:

1. How size of the instructor's writing
2. How much horizontal space the instructor is using for relevant content

Experienced lecturers tend to use half of a board at a time and minimize lateral motion, which permits the videographer to establish a static frame slightly wider than the relevant content. The instructor can move a few steps away from the content and remain within the frame.

When the instructor is projecting, attempt to establish a frame including the screen and at least the instructor's head. Most instructors will gesture at the screen and capturing those gestures is valuable. The zoom level will depend on each slide's legibility. If the instructor moves, the operator should use pan and zoom to balance the instructor and the screen in the frame.

30-minute video file limit

The EOS 80D stops recording after 30 minutes. The videographer must monitor the camera's timer and attempt to stop and restart the recording during a pause, ideally at a topical boundary, in order to provide a smooth transition for viewers. Cycle the recording any time after 25 minutes.

Common problems and solutions

The camera microphone's receiver offers no battery charge indicator. It therefore fails without warning. When it does, the status light will flash green. One pair of batteries typically lasts about two days.

When the batteries exhaust themselves, immediately unplug it and continue recording with the camera's built-in mic. Replace them as quickly and quietly as possible, then reconnect the mic. If the problem lies with the transmitter, do not stop the instructor: instead, address the issue after the session.

Maintaining session metadata

The videographer should maintain metadata about the sessions, namely:

• a count of the total number of sessions
• a session identifier in <day>.<session number> format
• the instructor's name and affiliation
• the lecture's topic
• the session's date, start time, and end time
• the card ID on which the session was recorded
• the length of each video file, in minutes
• notes about the recording: audio or video problems, etc.

Securely storing A/V equipment

We locked the A/V equipment in the venue's equipment room. The OPLSS coordinator was responsible for managing the key. The lead should acquire the key from the coordinator as early as possible before a session to retrieve the equipment.

Video Publishing Procedures

We published the videos to two locations:

• The University of Oregon's OPLSS site
• The OPLSS YouTube channel

Publishing to the OPLSS site

The OPLSS coordinator is the only person with credentials on the university's OPLSS site. Historically, the coordinator uploaded raw video files every evening. The videographer should give the day's data card to the coordinator at the conclusion of each day's sessions and retrieve the card from the coordinator before recording the next day.

Publishing to YouTube

Publishing the videos to YouTube is a complicated process involving multiple software components and a considerable amount of manual effort.

Creating title graphics

The videographer should create title graphics to accompany the session videos. We refer to them here as title cards or cards. We created one set of cards that appear in every video, along with a set of cards unique to each session.

The common cards are:

• main title
• sponsors
• copyright notice

The unique cards are:

• session title
• editor's notes

Each session card stated the topic, instructor, instructor's affiliation, lecture date, and session number on that date.

The main title, session, and sponsor cards appeared before the session video. The copyright card appeared after the video.

The videographer used GIMP to create the cards and xcf2png from xcftools to convert GIMP's XCF files to PNG format.

The OPLSS font is named PT Mono Bold. "PT" is a reference to ParaType, the font's creator.

The OPLSS font and logo use these colors:

Color	RGB (hex)
green	248c68
grey	828282
yellow	fee123

See the session schedule.

Generating the final videos

The videographer used FFmpeg and a set of custom tools to combine the raw video files and title graphics into files suitable for uploading to YouTube. The custom tools reside in this repository's bin directory. They're written in Ruby and require at least Ruby 2.4.

After generating the title cards, the videographer used FFmpeg to convert them to MPEG-4 videos, then concatenated the card videos with each session's video files.

The video elements appear in this sequence. The parenthesized number is the element's running time:

• main title card (6s)
• session title card (6s)
• sponsors card (6s)
• session video segments (~90m)
• copyright card (6s)

As noted earlier, the main, sponsors, and copyright cards were the same for all videos, whereas the session titles and recordings were unique to each session.

The custom tools partially automated the rendering process:

• from_image converted an image file into a short video with the same encoding as the session recordings. The lead used this tool to render title graphics as videos.
• sequence generated an ASCII sequence file in a format suitable for input to FFmpeg's concat. This file specified the video concatenation sequence.
• render used FFmpeg's concat and a sequence file to render a YouTube-compatible file. See also commentary on FFmpeg's concatenation options.

Due to the large file sizes and the time required to process the videos, the videographer should strive to maximally automate the video pipeline.

Storage considerations

The camera stored raw video files in MPEG-4 format on an SD card. This document refers to the SD card as source storage and the files as raw videos or source videos. These files typically range between 2 and 4 GB in size.

In addition to the SD cards, the videographer used a high-capacity portable hard disk to store processed videos and title cards. This document refers to this device as target storage.

Due to the large file sizes, the videographer first transcoded the source videos to target storage. Among other transformations described later, the transcoding process compressed the raw videos.

We recommend structuring the target storage file system as follows:

    assemblies/
      DD/
        SS.mp4
    titles/
        copyright.*
        main.*
        sponsors.*
      sessions/
        DD/
          SS.*
    segments/
      DD/
        SS/
          GG.mp4
    sequences/
        DD/
          SS.sequence

Definitions:

• assemblies contained the fully-assembled video files
• titles contained the title card static graphics and animated video files
• segments contained the session segment files after transcoding from source storage
• sequences contained the text-formatted sequence files that FFmpeg reads to assemble a final video from its constituent videos (animated title cards and session segments)
• DD was the zero-padded day number
• GG was the zero-padded segment number in a given session
• SS was the zero-padded session number on a given day

FFmpeg notes

FFmpeg is a powerful but notoriously complex tool. The videographer needed considerable trial and error to arrive at the options that follow.

Video processing invocations:

• To transcode videos from the camera's raw format to smaller files, use these options:

  -crf 30 -pix_fmt yuv420p -r 30 -vf scale=1280x720 -video_track_timescale 30k

  • -crf 30 compresses the resulting video
  • -pix_fmt yuv420p specifies a pixel format
  • -r 30 defines the frame rate (30/second)
  • -vf scale=1280x720 defines the resolution (1280x720)
  • -video_track_timescale 30k defines the "clock tick" timebase (30k ticks/second). The argument must match the animated title card's -- if not, the segment videos and cards will play at different rates

• Specify -ss h:m:s as an input option to skip the first hours:minutes:seconds of the video. Similarly, specify -t seconds as an input option to stop encoding after seconds. Use these options to excise unwanted content from a segment.

• To animate title cards, use these input options (i.e., options that appear before the -i option):

  -f lavfi -i anullsrc=r=48000:cl=mono -loop 1

  • -f lavfi specifies the input filter (lavfi)
  • -i anullsrc=r=48000:cl=mono inserts a silent audio track
  • -loop 1 loops the image as a video (1 loop per second)

• Likewise, use these output options (i.e., the options that appear after the -i option):

  -crf 30 -pix_fmt yuv420p -r 30 -t 6 -vcodec libx264 -vf scale=1280x720 -video_track_timescale 30k

  • -crf 30 same as transcoding
  • -pix_fmt yuv420p same as transcoding
  • -r 30 same as transcoding
  • -t 6 specifies the running time (6 seconds)
  • -vcodec libx264 specifies the video codec (H.264)
  • -vf scale=1280x720 same as transcoding
  • -video_track_timescale 30k same as transcoding

• To generate the sequence files, create a text file listing the constituents in the order by which to concatenate them, in this format:

  file 'PATH'
  .
  .
  .
  

  See the concat filter documentation for details.

• To assemble the final videos, use these options:

-safe 0 -f concat -i /sequence file/ -c copy /assembled file/.mp4

• -safe 0: suppress errors about unsafe file paths
• -f concat: specify the filter (concatenation)
• -i /sequence file/: specify the input file (sequence file path)
• -c copy: specify the output codec (copy)

Processing pipeline

The videographer processed the videos in the following sequence:

• Create an inventory:
  • Sessions, speakers, and topics
  • Raw video files and how they map to sessions
• Generate all of the title cards and animate them as MPEG-4 videos
• For each day:
  • For each session:
    • Transcode the raw video files from source storage to target storage
    • Generate an FFmpeg sequence file
    • Concatenate all of the session's constituent videos
• Upload the videos to YouTube

Uploading to YouTube

We published assembled session videos to the OPLSS YouTube channel. The videographer needs a Google account and "manager" permissions on the OPLSS brand account. See YouTube's brand account documentation.

YouTube recommends these encoding guidelines:

Parameter	Value
Container	MP4
Video codec	H.264
Audio	AAC-LC
Frame rate	Any standard rate
Bitrate	Varies by format
Resolution	Preferably full HD, i.e., 1080p / 1920x1080
Aspect ratio	16:9
Scan type	Progressive

See also YouTube's resolution suggestions.

A raw 90-minute 1920x1080 video may consume 8GB or more of storage, and therefore requires substantial upstream bandwidth to upload. The videographer used a wired connection in the U of O's computer science building to upload the videos. An 8GB file required about 7 minutes on that connection (20MB/s).