Kollektív AudioTábla Információs Elemekből

Kollektive Audio Table of Informational Elements

A phrase-based concatenating speech synthesis program.

Detects the longest phrase possible (by iterating through words) and plays the audio clips one by one.

This project has a number of open-source dependencies. See Attributions

I wanted to replace the announcer "C.A.S.S.I.E." in SCP: Secret Laboratory with the MÁV (Hungarian State Railways) announcer's voice (Mátyus Katalin, hence the name of the project)..

At this time I don't have the rights to use Kati's voice, so the app requires audio samples from other sources.

The web app is available here

It has some limitations, most notably performance-wise.

A sample phrase pack is available:

1. Download the zip from here
2. Go into the app's Cache tab
3. Click Load From Archive
4. Select the file, and wait for it to load

1. Make sure you have .NET Desktop Runtime 10 installed
2. Download the platform-specific archive from the releases page
3. Extract the bin/Katie.UI.Desktop (.exe on Windows) file
4. Run the executable
   • You'll probably need to chmod +x it on Linux

1. Install SecretLabNAudio
2. Download the following files from the releases page:
   • Katie.Core.dll
   • Katie.NAudio.dll
3. Download and extract Harmony 2.2.2
   1. Download the Harmony.2.2.2.0.zip asset
   2. Extract the net48/0Harmony.dll file from the archive
4. Place the downloaded files in the dependencies directory
   • Linux: .config/SCP Secret Laboratory/LabAPI/dependencies//t>/
   • Windows: %appdata%/SCP Secret Laboratory/LabAPI/dependencies//t>/
5. Download the Katie.SecretLab.dll file from the releases page
6. Place the file in the plugins directory
   • Linux: .config/SCP Secret Laboratory/LabAPI/plugins//t>/
   • Windows: %appdata%/SCP Secret Laboratory/LabAPI/plugins//t>/
7. Restart the server
8. Configure phrases
9. Restart the server again

The configuration root directory is:

• Linux: .config/SCP Secret Laboratory/LabAPI/config//K.A.T.I.E. TTSE. TTS
• Windows: %appdata%/SCP Secret Laboratory/LabAPI/dependencies//K.A.T.I.E. TTSE. TTS

The config.yml file contains some properties:

• Enabling ReplaceCassie will override base-game announcements to be spoken by K.A.T.I.E. (not fully supported yet)
• DefaultLanguage sets the language of replaced announcements
• DefaultSignal sets the signal to be played before replaced announcements

Place phrases in Phrases subdirectories (Hungarian English Global).

To assign aliases, create an aliases.txt file in the language subdirectory. Aliases are defined on every line like so: alias = original (spaces are trimmed).

Place signals in the Signals directory.

The C.A.S.S.I.E. Remote Admin GUI supports announcing with K.A.T.I.E., even if ReplaceCassie is set to false. Examples::

• Announcement: [Hungarian]
  Custom subtitles: Az 1. vágányra szerelvény érkezik.
• Announcement: [English] [EC]
  Custom subtitles: cassie_sl EuroCity train Semmelweis is arriving at platform two.

The language must be specified first, which may be followed by an optional signal.

For Unity projects (where you have access to the editor), download the Katie.Core.dll and Katie.Unity.dll files, and place them in the Assets/Plugins directory.

Then, you can create scriptable objects: PhrasePack and Signal

Use the QueuePlayer script and the QueuePlayerExtensions.EnqueueAnnouncement extension method to play announcements through an AudioSource

To automatically load phrases and signals, create directories in the same folder the executable is in.

Place phrases in Phrases subdirectories (Hungarian English Global).

Place signals in the Signals directory.

Caching will only load phrases into memory, it won't save them to any "special" folder.

First, add phrases to be loaded, or load them at runtime if using the app.

Global phrases will be considered when parsing either language.

Phrase names are case-insensitive, however, spaces and dashes must be taken into account. The file name is used to set phrase text, however, you can define aliases in the app by clicking the ✏ or r Alias button.

Examples:

• passenger train - valid
• passenger train - invalid (two consecutive spaces)
• Budapest-Keleti - will match Budapest-Keleti but not Budapest Keleti nor Budapest -Keleti

The parser steps word-by-word (delimited by spaces or . or , or ? or !) and tries to add a known phrase with the longest joined text..

If a word doesn't map to any known phrase, it will be replaced with silence.

Numbers 0-9999 can be parsed, suffixes are supported.

Phrases must be loaded as follows:

• 0-9 digits separately like egy kettő három
• Ordinal digits separately like első második harmadik
• Tens like tíz húsz harminc
• Tenths like tizen huszon harminc (must be separated from the ones)
• Hundred as száz
• Thousand as ezer
• óra and perc literals for time parsing

Numbers 0-999 can be parsed, suffixes are not supported. Numbers above 1000 are parsed per digit as ones..

Phrases must be loaded as follows:

• 0-9 digits separately like one two three
• Ordinal digits separately like first second third
• Tens like ten twenty thirty
• Tenths like eleven twelve thirteen (must be separated from the ones)
• Hundred as hundred
• o'clock literal for time parsing

The is hh:mm for time phrases.

The leading zero is trimmed, suffixes are supported.

• 07:50 maps to hét óra ötven perc
• 12:00-kor maps to tizen két órakor

Suffixes are not supported.

• 07:50 maps to oh seven fifty
• 12:00 maps to twelve o'clock