[Hands-on] Building a Real-Time AI Voice Bot

TODAY’S DAILY DOSE OF DATA SCIENCE

[Hands-on] Building a Real-Time AI Voice Bot

AssemblyAI​ has always been my go-to for building speech-driven AI applications.

It’s an AI transcription platform that provides state-of-the-art AI models for any task related to speech & audio understanding.

Today, let’s build the following real-time transcription apps with AssemblyAI.

The workflow is depicted below, and the video above shows the final outcome.

• ​AssemblyAI for transcribing speech input from the user [speech-to-text].
• OpenAI for generating response [text-to-text].
• ElevenLabs for speech generation [text-to-speech].

The entire code is available here: ​Voice Bot Demo GitHub​.

Let’s build the app!

VOICE BOT APP

Prerequisites and Imports​

Start by installing the following libraries:

Next, create a file app.py and import the following libraries:

VOICE BOT APP

Implementation

Next, we define a class, initialize the clients involved in our app—AssemblyAI, OpenAI, and ElevenLabs, and the interaction history:

• ​Get the API key for AssemblyAI here →​
• ​Get the API key for OpenAI here →​
• ​Get the API key for ElevenLabs here →​

Now think about the logical steps we would need to implement in this app:

1. The voice bot will speak and introduce itself.
2. Then, the user will speak, and AssemblyAI will transcribe it in real-time.
3. The transcribed input will be sent to OpenAI to generate a text response.
4. ElevenLabs will then verbalize the response.
5. Back to Step 2, all while maintaining the interaction history in the self.interaction object defined in the __init__ method so that OpenAI has the entire context while producing a response in Step 3.

Thus, we need at least four more methods in AI_Assistant class:

• generate_audio → Accepts some text and uses ElevenLabs to verbalize it:

• generate_ai_response → Accepts the transcribed input, adds it to the interaction, and sends it to OpenAI to produce a response. Finally, it should pass this response to the generate_audio method:

• start_transcription → Starts the microphone to record audio and transcribe it in real-time with AssemblyAI:

1. on_data → the method to invoke upon receiving a transcript from AssemblyAI. Here, we invoke the generate_ai_response method:
2. on_error → the method to invoke in case of an error (you can also reinvoke the start_transcription method).
3. on_open → the method to invoke when a connection has been established with AssemblyAI.
4. on_close → the method to invoke when closing a connection.
5. These three methods are implemented below:

• Lastly, we have stop_transcription → Stops the microphone and let the OpenAI generate a response using the method below.

Done!

With that, we have implemented the class.

VOICE BOT APP

Demo

Finally, we instantiate an object of this class and start the app:

Done!

This produces the output shown in the video below:

That was simple, wasn’t it?

You can find all the code and instructions to run in this GitHub repo: ​Voice Bot Demo GitHub​.

VOICE BOT APP

A departing note

I first used ​AssemblyAI​ two years ago, and in my experience, it has the most developer-friendly and intuitive SDKs to integrate speech AI into applications.

​AssemblyAI​ first trained Universal-1 on 12.5 million hours of audio, outperforming every other model in the industry (from Google, OpenAI, etc.) across 15+ languages.

Now, they released ​Universal-2​, their most advanced speech-to-text model yet.

Here’s how ​Universal-2​ compares with Universal-1:

• 24% improvement in proper nouns recognition
• 21% improvement in alphanumeric accuracy
• 15% better text formatting

Its performance compared to other popular models in the industry is shown below:

Isn’t that impressive?

I love ​AssemblyAI’s​ mission of supporting developers in building next-gen voice applications in the simplest and most effective way possible.

They have already made a big dent in speech technology, and I’m eager to see how they continue from here.

Get started with:

Their API docs are available here if you want to explore their services: ​AssemblyAI API docs​.

🙌 Also, a big thanks to ​AssemblyAI​, who very kindly partnered with us on this post and let us use their industry-leading AI transcription services.

👉 Over to you: What would you use ​AssemblyAI​ for?