Word Salad v1.1.0

The What

"Word Salad" is a text-to-speech application with a twist: rather than using speech synthesis, it instead uses sentence mixing. By providing pre-recorded words as individual files, "Word Salad" is able to assemble sentences to play back as audio. Perfect for Twitch redeems and the like!

The Why

Text-to-speech is a common redeem that I see on many channels. However, I feel that the open-ended nature encourages chatters to say whatever intrusive thought crosses their mind, with little concern for length or disruption. By restricting the pool of available words, "Word Salad" presents the chatter with a fun puzzle. Thus, to say anything coherent at all feels like a small victory, and figuring out how to convey your desired sentiment requires critical thinking. It's good exercise for the mind and soul.

The How

This application is powered by Node.js in an Electron shell for the frontend.

If you're running this via an executable file, you can skip this section.

If you're building this from source, you will need:

• Node.js
• npm

From there, fire up terminal and run good old npm install. Then, you can launch the application with npm start. This should fire up Electron and open a window with the UI.

You can also package it into a portable executable file yourself by running npm run build, which will output into the dist folder. This process uses electron.build.

Adding Words

Obviously, I can't ship this application pre-filled with audio that I don't own, so you'll have to create your own Word Banks. To do this, follow these steps:

1. Click the green plus icon next to the World Banks dropdown to create a new Word Bank.
   • You can also just rename the default, empty Word Bank.
2. Select your new Word Bank from the dropdown if it isn't already.
   • Important! Observe that your bank has a display name, and also Bank ID. This ID will be formatted like 5ea7fc19-5be6-4d62-85d4-1ef80f8fdeb7. You will need this ID later when making Speak commands.
   • You can rename your Word Bank at any time, but the ID will remain constant.
3. Create a folder and populate it with audio files.
   • Accepted file formats are .wav, .mp3 and .mp4.
   • File names must be formatted as word_number.wav. The filename informs the application what each word is. For example, the_02.wav represents a variant of the word "the".
   • Optional! You can designate sounds to play for words that are not present in the bank by simply naming the sound file _. For example, _.wav and _1.wav could be "Umm" and "hmm". If these files are not present, the player will simply skip over unrecognized words.
   • Subfolders within main folder folder are allowed.
4. Click the "Select File Folder" button to navigate to the folder where your audio files are stored. This should automatically populate your Word List.

As a quick example, you can download all of the Half-Life 1 VOX word files from here, or from the game install if you own the game itself. These files are even already named in the correct format, which is very convenient!

If you're looking for a good jumping-off point for choosing your own words from your own source, I found the Wikipeda List of the 100 most common words in the English Language to be a useful guide in determining which words were most important to find and add to the Word Bank from my chosen source.

Making It Speak

The application has two components: the Main App and the Browser Source. Once you have populated your Word Bank folder as outlined in the previous section, you can make the application speak by following these steps:

1. Add the speaker as a Browser Source to your OBS Scene with the URL http://localhost:8095/speaker. Be sure to check the box allowing OBS to control the Audio of this source!
   • Important! 8095 is the default port. If you change the Port Number in your Settings, be sure to update the URLs of your Browser Sources.
   • If you start OBS before starting this application in the future, you may need to refresh your Browser Source. Without the Main App running first, the Browser Source can't be served.
   • It is recommended to go into "Advanced Audio Settings" and enable "Monitor and Output" for this Browser Soruce's audio in OBS.
2. Issue a speak command to the Main App by sending a GET request to http://localhost:8095/speak?bank=<Bank ID>&phrase=<your sentence here>.
   • For example, with a Word Bank with Bank ID 5ea7fc19-5be6-4d62-85d4-1ef80f8fdeb7 and the desired phrase You dense Diglett!, your request would be http://localhost:8095/speak?bank=5ea7fc19-5be6-4d62-85d4-1ef80f8fdeb7&phrase=You dense Diglett!
   • You can hit this URL in any browser to issue the command, or use the Main App's UI.
   • Optional! Speak commands will be processed in a queue, in order to prevent the system from speaking over itself. If you would like to disable this behavior, add ?queue=false to the end of the Browser Source URL, like http://localhost:8095/speaker?queue=false.
3. If you had Audio Monitor enabled, you should have heard the phrase play, provided that the words you wanted to say exist in your Word Bank!

Setting Up Twitch Redeems

The current version of "Word Salad" doesn't actually feature any direct Twitch integration at all! Instead, you can use it in tandem with a Channel Manager like MixItUp by following these steps:

1. Create a Channel Point Redeem via the Twitch Dashboard.
   • Make sure to check the "Require Viewer to Enter Text" box.
2. In MixItUp, edit the new Channel Point redeem by adding an action that issues a Web Request to http://localhost:8095/speak?bank=<Bank ID>&phrase=$message.
   • The $message variable in MixItUp passes the content of the redeem's text to the Web Request.
3. Get the list of words your chatters can use by going to http://localhost:8095/banks/<Bank ID>/words, or by clicking the "Copy" button next to your Word List on the Man App's UI. It is recommended that you take this list and put it somewhere publicly accessible. Your homepage, your Twitch bio, wherever works!

You can set up different redeems for different Word Banks by repeating the above steps with different Bank IDs!

Troubleshooting

This app was envisioned and constructed in about two weeks by Tom "Skeletom" Farro. It's entirely possible that something doesn't work, or there's a setup step he forgot to document. Please feel free to reach out to him on social media or via email if you encounter any issues!