softchat

• convert twitch/youtube chat into softsubs (webm)
• supports windows, macos, linux, freebsd
• MIT-Licensed, 2020-10-12, ed @ irc.rizon.net

quality

• well-tested, well-regarded software
• check out some of the authentic user feedback below

readme TOC

• top
• howto
  • first-time setup
  • regular usage
  • checking for updates
  • common problems
    • certificate errors on windows
    • missing cookies
• advice
• bugs
• about
• dependencies
• maintainers

howto

windows users:

• press Win+R and run cmd to get a terminal to run commands in (protip: rightclick to paste stuff)
• replace python3 with python in all commands
• see common problems if you run into issues

first-time setup

• run these commands:

  python3 -m pip install --user softchat chat-downloader fugashi[unidic]
  python3 -m unidic download
  

• download https://ocv.me/stuff/NotoSansCJKjp-Regular.otf into a folder named noto-hinted in the same place you keep your VOD rips

• if you want graphical emotes, see --emote_font in dependencies

regular usage

• download/convert a youtube VOD automatically with ytdl-tui.py (save it next to the noto-hinted folder, not inside it), then either doubleclick it or:

  python3 ytdl-tui.py youtube.com/watch?v=4dlsbJiEa-U
  

• ...or do the downloads/conversion steps manually like this:

  python3 -m youtube_dl youtube.com/watch?v=4dlsbJiEa-U
  python3 -m chat_downloader --message_groups all -o same-filename.json youtube.com/watch?v=4dlsbJiEa-U
  python3 -m softchat -m2 same-filename.json
  

  the json filename should be whatever youtube-dl decides to use, except replace the extension with .json (and remember quotes depending on OS)

checking for updates

update to the latest version by running this:

python3 -m pip install --user -U softchat chat-downloader

this does not cover ytdl-tui; to check for ytdl-tui updates, run it and press uu

common problems

certificate errors on windows

errors like the following:
[SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed: unable to get local issuer certificate

• open Internet Explorer (yes really) and access each of these URLs once:
  • https://pypi.org/ (to install stuff with pip)
  • https://raw.githubusercontent.com/ (if you use ytdl-tui)
  • https://youtube.com/ (no need to accept cookies or consent to whatever, just accessing it is enough)
• _{apparently this is how windows loads certificates into the cert-store? ms pls}

missing cookies

symptoms:

• youtube-dl failing because the video is members-only
• chat-downloader spamming a bunch of javascript then exiting

note: generally ytdl-tui should handle most of these scenarios, so you could try that first, otherwise read on

let's export your cookies and make them available to chat-downloader and youtube-dl; should dodge captchas, make it possible to donwload members-only stuff, and also fix the recent consent requirement from EU members

• install the following addon: https://addons.mozilla.org/en-US/firefox/addon/cookies-txt/
• access youtube.com, click the extension icon, [Current Site]
• save as cookies.txt into your VODs folder (next to the noto-hinted folder)
• ytdl-tui will now avoid the issue

repeat these steps if it breaks again

if you run chat-downloader manually instead of ytdl-tui, append these arguments: -c cookies.txt

.

advice

mpv is the recommended player, however in danmaku mode, you should resample the video to your monitor's FPS.

• one way to do this: adding the argument --vf=fps=120 every time
• another way is adding the following line to input.conf, which lets you press shift-U to toggle upscaling to 120 fps:

  U vf toggle fps=120
  

• on linux/mac, input.conf should be here: ~/.config/mpv/input.conf
• on windows, input.conf should be here: %appdata%\mpv\input.conf

additionally, on windows you may need to --vo=direct3d and not fullscreen for resampling to work -- try the different alternatives in --vo=help and find one which:

1. doesn't ruin the contrast
2. doesn't make the text blurry
3. doesn't make the text jitter

The fonts produced by --emote_font need to be somewhere your media player can see them (either embedded, or in some autoload folder, or installed as system fonts). Set --emote_install to install fonts into ~/.config/mpv/fonts/ (Linux/macos) or %appdata%\mpv\fonts (windows) for local playback using mpv. For other media players, or if you intend to share the subtitles, you should use --embed_files as each font is specific to that particular subtitle file and they need to be used together.

Override automatically vectorized emotes by creating a ".manual.svg" file in the cache directory. So for the emote UCS9uQI-jC3DE0L4IpXyvr6w_Do6dXuL6LZCQ_AP4-buoDQ create the file UCS9uQI-jC3DE0L4IpXyvr6w_Do6dXuL6LZCQ_AP4-buoDQ.manual.svg. This can be used for emotes where the automatic version is of low quality.

Similarly, rather than using --emote_fill to fill the background of all emotes, you can create an empty file named UCS9uQI-jC3DE0L4IpXyvr6w_Do6dXuL6LZCQ_AP4-buoDQ.bg to enable bg-fill for just that one emote.

bugs

• --emote_sz 3 and higher makes things collide (wontfix probably)

about

• superchats will display for 2x the time and with inverted colors

• moderator messages are emphasized
  (larger outline, and prefixed with a ball)

• mode 1, sidebar chat, creates a huge amount of subtitle events
  which many media players (including mpv) will struggle with

  for example, seeking will take like 5sec

  you can fix this by muxing the subtitle into the vid:
  ffmpeg -i the.webm -i the.ass -c copy muxed.mkv

• mode 2, danmaku, will look blurry and/or jerky in most players
  unless you have the subtitles render at your native screen res

  for example in mpv you could add these arguments:
  --vo=direct3d --sub-delay=-2 --vf=fps=90

  replace 90 with your monitor's fps

• after an upgrade, you can reconvert old rips like this:
  grep -lE '^Title: .*softchat' -- *.ass | tr '\n' '\0' | xargs -0rtl python3 -m softchat -m2 --

• youtube VOD chatlogs are incomplete (about 80% of messages are lost)
  so softchat can now take multiple chat JSONs to splice together:
  it is recommended to run chat_replay_downloader.py twice,
  first when the stream is live and then afterwards for the VOD chat;
  when running softchat, the VOD json should be the first file provided,
  followed by any live recordings to splice messages from

• on windows, --kana requires python 3.8 or newer

• on macos, --kana is currently busted in macports on sonoma

dependencies

• chatrips must be made using https://github.com/xenova/chat-downloader

  • softchat will be updated as the output format of chat-downloader changes

• a copy of NotoSansCJKjp-Regular.otf in a folder named noto-hinted

  • download here: https://ocv.me/stuff/NotoSansCJKjp-Regular.otf
  • or here: https://noto-website-2.storage.googleapis.com/pkgs/Noto-hinted.zip
  • in the future (tm) softchat will require all the Noto-fonts in this folder

• required python libraries: Pillow fontTools

• enable --kana (convert kanji to hiragana) with these:

  python3 -m pip install --user fugashi[unidic]
  python3 -m unidic download
  

• enable --emote_font with:

  • fontforge
  • ImageMagick
  • potrace (not required on windows)
  • on debian: apt install potrace python3-fontforge
  • on macos: port install imagemagick potrace fontforge

note: on windows, cmd must be reopened after installing dependencies

maintainers

• build and release: python setup.py rls