From 7b48f8e6cc81df20c9c1b8c4bb98e3b1c2fc26f8 Mon Sep 17 00:00:00 2001 From: Sacha Chua Date: Sat, 10 Dec 2022 08:43:31 -0500 Subject: add harvesting tips --- harvesting.md | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 93 insertions(+) create mode 100644 harvesting.md diff --git a/harvesting.md b/harvesting.md new file mode 100644 index 00000000..89b6f006 --- /dev/null +++ b/harvesting.md @@ -0,0 +1,93 @@ +[[!meta title="Harvesting Q&A]] +[[!meta copyright="Copyright © 2021, 2022 Sacha Chua"]] + +During the harvesting phase of the conference, we work on collecting +the ideas that people shared in the Q&A sessions as well as any talks +that were not available as pre-recorded videos. It's a great way to +help speakers get stuff out of their heads and into a form we can all +learn from. Here's a process for doing so. + +# Skim the transcript and the videos to see if anything needs to be removed, and which video to use + +BigBlueButton gives us the webcams and audio as one video +(`--bbb-webcams.webm`) and the screenshare (if any) as another video +(`--bbb-deskshare.webm`). If the speaker shared their screen, we can +focus on that instead of their webcam. The following ffmpeg command +combines the audio from the webcams (which has been previously +extracted into a separate file, `--bbb-webcams.opus`) with the video +from the screenshare. + + ffmpeg -i example--bbb-webcams.opus -i example--bbb-deskshare.webm -c copy example--answers.webm + +We also want to check if people accidentally shared sensitive +information on their screen, or if anyone said something that they +might not have said if they remembered that thethe Q&A videos will be +shared after the talk. Sometimes there's some time before we get +around to closing the meeting at the end of the Q&A. Usually, a quick +read of the transcript will show anything that needs to be trimmed. +Here's how to stop the recording at a specified time: + + ffmpeg -i input.webm -to hh:mm:ss -c copy output.webm + +Cutting out stuff from the middle of a recording is slightly more +complicated. It might be easier to use a nonlinear video editor such +as kdenlive to edit the video. If you want to use ffmpeg, using +filters to select the frames and reencode the video will probably work +out better than splitting the file into multiple parts and then +concatenating them without reencoding, as the latter tends to need to +be split on keyframes. Here's a sample command based on this +[StackOverflow](https://stackoverflow.com/questions/64866231/remove-a-section-from-the-middle-of-a-video-without-concat) answer that removes the section between 15 seconds and +45 seconds: + + ffmpeg -i input.webm \ + -vf "select='not(between(t,15,45))', setpts=N/FRAME_RATE/TB" \ + -af "aselect='not(between(t,15,45))', asetpts=N/SR/TB" \ + output.webm + +Alternatively, you can let us know what parts needs to be trimmed, and +we can figure that part out. + + +# Add chapter markers + +Chapter markers make it easier for people to jump to the part of the +Q&A that they're interested in. You can see an example in the [Q&A for +asmblox](https://emacsconf.org/2022/talks/asmblox/). You can make a text file with the hh:mm:ss or mm:ss +timestamps and the chapter headings. + + 00:00 Introduction + 01:12 Why did you choose an internal state versus many 'state buffers'? + 02:10 Do you have plans to port shenzhen.io to Emacs? + 02:29 Did this use WASM? + 02:59 Why wasm rather than a more traditional Assembly dialect? It wouldn't be harder to implement, right? + 05:08 Any next projects on your mind? + 05:52 Does this work with any other paren-based editing packages? + 06:46 What kind of tool could use this idea? + 07:56 How did you go about designing the puzzles? + 08:39 What are your favorite changes in the upcoming Emacs 29? + 09:07 Are there tools to add more puzzles? + +Alternatively, you can edit the VTT file (`--bbb-webcams.vtt`) and add +NOTE comments with the chapter headings before the subtitles that are +part of that chapter. If you're using [subed](https://github.com/sachac/subed) to edit subtitles within +Emacs, you can split the subtitle as needed with `M-.` +(`subed-split-subtitle`) so that the subtitle starts with the +question. You don't have to worry about getting the timestamps exact, +as we can re-align them with `M-x subed-align`. Here's what that NOTE +comment can look like: + + NOTE Why did you choose an internal state versus many 'state buffers'? + + 00:01:12.600 --> 00:01:16.039 + Okay. So, the first question is why did you choose an internal state + +These can then be extracted with +`emacsconf-subed-make-chapter-file-based-on-comments` from +and included in our publishing workflow. + + +# Edit the transcript + +If you want to make it even easier for people to learn from the Q&A, +you can edit the transcript so that it can also be published on the +wiki page. See [[Captioning tips|captioning]]. -- cgit v1.2.3