/where-do-we-begin

Primary LanguagePython

where do we begin

files

  • download.py - script for downloading podcast audio and cover images, then calling a script to merge them with ffmpeg.
  • etc.

usage

scripts

Example use:

# create the file urls.txt
python download.py
# download the urls in urls.txt
./download-wget.sh
# merge audio and images into video
# this uses merge-test.sh which you can also run manually
python download.py
# at this point upload the video to youtube
# upload the transcript
# and download the resulting aligned subtitles
# create label tracks from subtitles
python separate.py

audacity

working with the labelled audio use Edit > Labelled Audio > Split Cut https://manual.audacityteam.org/man/edit_menu_labeled_audio.html

for some reason I'm getting an error "There is not enough room available to paste the selection" not trying to paste just trying to cut https://manual.audacityteam.org/man/error_insufficient_space_in_track.html "Split delete" works fine.

I could duplicate the main track for each speaker then split delte the inverse? seems impractical

ok it works if you do toggle Tracks > Sync-Lock Audio to ON you might wanna mix down to mono to speed things up too first mark out the first time each speaker talks, for offset. copy-paste the label track or whatever

  1. select the source audio track and one of the label track
  2. create a new stereo track and solo it
  3. split cut from labelled audio
  4. paste into a new track. make sure to select the point at which it starts, to get the right offset
  5. immediately export, then clear the out-track, to avoid killing your pc
  6. finally, export what remains of the original track

mix and render doesn't preserve start offset either

todo make a macro for this

ep. 1 is 01:42:39 and the intro is 24.019s

todo

could add infobox to each podcast ep. preceding, number, sers etc.

links

windows xp mode

ISO

activation key

virtualbox stuff

windows media player

notes

the basic display resolution for xp is 800x600

visualisers

Ambient does NOT go fullscreen. this is not a driver issue or anything, it's just htat visualiser OLDEST

  • Ambience
  • Plenoptic
  • Battery
  • Alchemy NEWEST

1920/3 = 640 1080/2 = 540 640x540 obs transform shows 1280 right, 540 bottom

gonna use guest size 1024x768

crop offset 192, 114

Base -- Bars rest alphabetical, vote later Baaulp - Alchemy Gir -- Ambience Log -- Battery Trog -- Plenoptic Wayne - Alchemy

mm not enough for each unique all alchemy?

how to record multiple windows at once?

  1. multiple vms - won't be able to sync them all
  2. recording one-by-one - seems to be the main option
    • can do it unattended by putting them in a playlist
  3. recording with virtualbox - kills performance
  4. multiple obs instances - is this possible?
  5. in-vm recorder - might be possible. have to find an old download
  6. multiple vms, record all at once then sync in edit
  7. see what Qemu offers
  8. write a script to syncronise it
    • could do this with a virtualbox interface, or with batch scripting over the local network
  9. running them all in one vm - can't work, need fullscreen or crop. could fix it with editing too like 6.

do this tomorrow, don't have time to run it today.

could play at 2x speed 60fps and halve it to 30fps in edit. I dont' think this is worthwhile.

lovely how windows xp mute doesn't affect the login sound even with single user. I guess there's a settings for that good thing vbox remembers window size. of course, I'm sure if it didn't there would just be yet another method/workaround could disable effects etc. to improve performance?

start wmplayer "C:\Documents and Settings\Joel\My Documents\My Music\1\Base.wma" /fullscreen

remove mouse first have to run this twice for it to fullscreen right disable mouse capture in vbox and in obs first and in tom and jerry lol!!

how to do it:

  1. create a shortcut to Z: (shared folder) and a shortcut to "C:\Documents and Settings\Joel\Start Menu\Programs\Startup" for convenience
  2. create a vm clone for each speaker
  3. place the vm's respective script in the startup folder.
    • better yet, place a shortcut to the script in the shared fodler, so you can edit the script without updating it in the vm
    • if you don't use a shared folder you could use some vbox script to edit the vhd file directly. to update scripts
  4. now that vm will play that speaker on startup! and using the shared folder you'll never have to edit that vm
  5. start all the clones at once, get the windows set up on the desktop and in obs
  6. reset them all to sync

VBoxManage setextradata global GUI/SuppressMessages confirmGoingFullscreen,remindAboutMouseIntegration,remindAboutAutoCapture see https://forums.virtualbox.org/viewtopic.php?f=7&t=107743

https://static.wikia.nocookie.net/marvelcinematicuniverse/images/6/64/Fury_and_the_WSC.png/revision/latest/scale-to-width-down/1000?cb=20160228230156

keyboard shortcuts

CTRL+P play CTRL+S stp CTRL+B back CTRL+F Forwards CTRL+SHIFT+B rewind ALT+ENTER Fullscreen

Host+A adjust window size

next up: always-on livestream of all wdwb episodes with windows media player visualiser

was thinking of running it on the vm live, realised if I want to do the illuminati council thing I should pre-record it. for a single virtualiser I can run it live. take a poll what people like more?

no ui (api only) 1,000 free minutes per month (16 hours!) https://symbl.ai/pricing/ https://docs.symbl.ai/docs/async-speaker-separation

no pretrained https://github.com/facebookresearch/svoice https://github.com/mindslab-ai/voicefilter costs money like £15 per hour of audio https://www.simonsaysai.com/help/4176715-automatic-speaker-identification

uploads

try aligning transcription to subtitles with youtbe https://wayneradiotv.fandom.com/wiki/Where_Do_We_Begin%3F_Episode_1:_The_Future_Of_Podcasting

channel I namecamped https://www.twitch.tv/wheredowebegin

episodes

NEED to record on obs with more bitrate. the fucking visualisers are huge bitrate eaters

todo [ ] download all episodes. write a script to download them then attach the associated image to make video. may use lots of disk space [ ] make transcripts into subtitles [ ] run livesteam with visualiser [ ] (maybe) figure out how to split speaker, make the illuminati real

AHAH: 0. I could search for something to segment audio by speaker. splitting crosstalk is not required. 2. REAL: I can use the transcripts! they're already captioned by name. I need to align the transcripts to timed text already anyway, for subtitles. eureeeek.

Title:

24/7 where do we begin podcast with visualiser. rtvs podcast (wayneradiotv, baaulp, trog, log, gir) tags

I should ask rtvs by podcast email "if you were a wmp visualiser which one would you be" and ask permission to run the stream at the same time

could fork podcastparser to make it return a typed dict

I have to do all the podcast uploads to yt with still images before making visualisers, cause I need it to align the subs to split the tracks. UNLESS I can find a local tool to do it?? for splitting, I should get my program to just make an audacity label track. that simplifies things, then I do the hard part manually in audacity so I don’t have to fucking write it. not free software https://www.captionhub.com/blog-post/automatically-align-transcript “video transcript alignment” only one github star, python2.7 + perl :/// https://github.com/polizoto/align_transcript https://pypi.org/project/subaligner/ here we go https://github.com/baxtree/subaligner https://github.com/amiaopensource/An_Archivists_Guide_To_Matroska/blob/master/metadata.md#adding-tags-with-ffmpeg