Latest news

From Command Bots to AI Bots: a Twitch Botosphere Case Study

No comments

From Command Bots to AI Bots

A Twitch Botosphere Case Study

Bots, bots, bots, bots: the word has a different echo around the internet. On Messenger, you get cheers. On Twitter, you get anti-spam protests. On Discord, you think of mechanical command bots. Today, we feel the time is right to create a new echo for AI bots on Twitch, the n°1 gaming community in the world.

Many bots already exist on Twitch and are well integrated. They usually provide important information to stream viewers or manage the moderation of chat rooms. Sometimes, they’re used for doomer purposes. But if bots are still used daily, it means even in their current basic form, they bring something to the Twitch community.

Most Twitch bots are currently very basic: command activated, they send automatic messages when triggered. The community is open to them: they’re seen as useful, even if sometimes a bit boring or unpractical. How could they evolve to further transform the world of streaming?

Three words: understand human language.

The power of truly understanding human language

I hear many people from all around : “but current bots are just fine! Why would we change them?”. To answer that, I want to talk about 3 things that have proven their values: user experience, performance, and multilingualism.

Building viewer loyalty through a fantastic user experience

List of commands for a command bot

As I’m writing, a few Twitch users will be typing !uptime to know how long the streaming they just joined has been going on for. Many Twitch users don’t know the command to get that information, or any other command for that matter. That makes current bots hard to reach and very, well… robotic.

Creating bots with NLP (natural language processing) solves this issue very simply: it makes them understand human language in addition to commands. Experienced viewers could still type !uptime, but new ones could simply say “how long have you been streaming?” and get the same answer. The rules of banning could be more natural, meaning people would stop getting banned for using curse words without insulting anyone. The general chats would not be overwhelmed with basic questions but gather interesting and fun content. I could go on!

Giving Twitch bots the power of understanding human language brings us the simplicity of conversing with them like we would with a friend. Imagine the things you could do!

Impressing viewers through deep understanding

NLP is a subtopic of artificial intelligence. By using AI, you change your approach to language understanding. Bots don’t base themselves on keywords, they understand the global meaning of a sentence. That makes them smoother and more adaptable, and in the end, provides a more natural experience for the user.

“But what if the bot doesn’t understand what the user said? That happens.”

The beauty of machine learning is that bots are always learning. It doesn’t understand once? It’ll understand the next time. And for things the bot doesn’t support, bot masters (whether a streamer making a bot for their channel or a bot-addicted viewer) can take on these questions themselves. That’s what we call human fallback. So from the viewers’ point of view, their questions never go unanswered.

Reaching out to international viewers with multilingual support

It’s important to care for your existing community. Twitch, like the majority of the internet, is English-speaking. But many viewers aren’t native English speakers, and are craving for content in their own language. Multilingual bots can understand any language and provide answers in the user’s tongue, making the viewers’ experience smoother and closer to home. Ideal to reach out to new viewers, or treat existing ones.

And what’s in it for the streamer? A multilingual bot manages all basic questions for them, so the streamer can focus on animating a fun and insightful conversation about the stream itself. Nothing else.

Changing the bot landscape in the streaming world

Twitch, as a social platform and a messaging channel, offers endless possibilities for bots. With both private messages and general chats, there are great ideas all around. Here are a few we gathered:

Stream general information

When discovering a new stream, you can easily get confused: what’s going on? What is this game, which version is it? How long has this stream been going on? When does it end? A bot is the perfect way to provide precise answers to these questions in a very natural ways.

Streamer general information

What is this streamer’s setup? Where are they from? Do they often compete? What mic are they using? Many questions can be answered by a bot integrated into the streamer’s channel.

Streaming schedule

Major streamers typically have a webpage where people can access their schedule to discover if and when they’re going to play a certain game. Most share the information on social networks. Including this information directly in Twitch and so keeping the viewer on the stream, is a good way to increase viewing rates. A bot can easily provide the information to anyone in natural language.


Ever wanted instant access to all streamer rankings on Twitch? Make the ultimate Twitch ranking bot!

Smart moderation

Instead of banning people for specific keywords, why not use AI to provide sharper moderation? Language technologies allow the analysis of text and can recognize inappropriate sentences or paragraphs, even when they don’t use key swear words. That way, you wouldn’t get banned for using swear words in an inoffensive way.

Chat insights

Hard to follow the conversation when loads of people are chatting at the same time? It’s easy to use Recast.AI to analyse the most talked about topics and provide answers to the most frequent questions. Great way to connect a community.

We could go on and on, but I’ll leave this in the hands of the very creative and capable Twitch community! And if you are a bot builder, send us your Twitch bots, we’ll be happy to promote them 🙂

Building your own Twitch bot

“That’s all very well, but how can I do this?”

Recast.AI allows you to easily build and connect bots to Twitch through a multi-platform tool called Bot Connector. It’s as easy as it gets! You’ll find below a detailed process on how to launch your own bot on Twitch. Enjoy!

Create a bot on Recast.AI

Go to Recast.AI, create an account and start building your bot. Here’s a tutorial to help you get started on Recast.AI.

Create an account on Twitch

Go on Twitch and create an account. If you already have a Twitch account, login.

Get a Client ID

Go on the connections page of your Twitch account and register a developer application. Scroll to the end of the page and click on Register your application.

Fill in the form on the Recast.AI Connector page. You can fill in the Redirect URL with a random url, as you don’t need it. Select Chat Bot in the Application Category and click on Register to create your app.

Fill in the Recast.AI form with your Twitch username and with the Client ID of your app.

Get an oAuth token

Go on the Twitch Password Generator and click on Connect with Twitch

Copy the password generated in the Token field on the Recast.AI platform (excluding the oauth:part)


Verify all the information provided on the Recast.AI fields is correct and click on Connect! Your bot is now live.

If you need any assistance, don’t hesitate to join our Slack Community, our support team will be happy to help. If you have other ideas, want to discuss this material, or have built a bot, please leave a comment!

From Command Bots to AI Bots: a Twitch Botosphere Case Study was originally published in Chatbots Magazine on Medium, where people are continuing the conversation by highlighting and responding to this story.

Source: Chatbots Magazine

magnoliaFrom Command Bots to AI Bots: a Twitch Botosphere Case Study

Related Posts

Leave a Reply