Getting Started
Create your first AI voice pipeline in minutes.
Learn how to create your first voice pipeline for real-time conversational AI. This guide will walk you through logging in, creating a pipeline, and testing it in our playground.
Sign Up and Login
- Visit dash.layercode.com
- Sign up or log in using email and password, then verify your email.
- You’ll be directed to your dashboard where you can manage your pipelines.
Configure Your Voice Pipeline
After logging in for the first time, you’ll be redirected to your first pipeline, created from our recommendedtemplate.
Pipelines can be customized through an intuitive UI with settings for transcription, text-to-speech, and backend (which generates the AI’s response to be spoken). Click the “Edit” button to on any box in the pipeline to configure it.
Feel free to leave all the default settings as is, and skip to testing your pipeline below.
Let’s take a look at the settings available for each stage of the pipeline:
Transcription Settings
Configure how user speech is converted to text. The default transcription provider and model are optimized for low-latency English language transcription. For multi-language support, there are specialized transcription models we support.
| Setting | Description | Options / Notes |
|---|---|---|
| Provider | The transcription provider used for speech-to-text. | See available transcription providers |
| Model | Select the transcription model that best fits your needs. | See available transcription models |
| Turn Taking Mode | Determines how the system detects when the user is speaking. | Automatic: AI detects when user has finished speaking (using silence detection). Push to Talk: User controls when they’re speaking by holding down a button. Read more about turn taking. |
| Can Interrupt | When Turn Taking Mode is ‘Automatic’: Toggle whether the user can interrupt AI whilst it’s speaking | Enable or disable interruption. When disabled, the user can only respond once the AI has finished speaking. |
Click “Save Changes” to apply your transcription settings.
Text-to-Speech Settings
Configure the text-to-speech provider and model used to turn text generated by the backend into speech spoken by the AI. The default provider and model are optimized for low-latency English language use cases. For multi-language support, there are specialized text-to-speech models we support.
We recommend experimenting with different providers, models and voices as they all have varying characteristics.
| Setting | Description | Options / Notes |
|---|---|---|
| Provider | The text-to-speech provider used to generate AI speech. | See available text-to-speech providers |
| Model | Select the TTS model that matches your quality and speed needs. The default model is often best for English language use cases. | See available text-to-speech models |
| Voice | Choose the voice that best represents your AI. | Select from available voices in the chosen provider/model. We recommend experimenting with different voices to find the right one for your use case. |
Click “Save Changes” to apply your text-to-speech settings.
Backend Configuration
The Backend receives the transcribed user’s speech, and is responsible for generating the voice AI’s response. Layercode offers a hosted backend or the ability to connect your own backend with a simple webhook.
Get started immediately with Layercode’s optimized backend powered by Gemini Flash 2.0. Our hosted backend provides:
- Ultra-low latency responses
- Optimized for real-time conversation
- Zero backend setup required
| Setting | Description |
|---|---|
| LLM Prompt | Configure the personality and behavior of your AI assistant. |
| Welcome Message | Configure the message your AI will speak when the conversation first starts. If disabled, the user starts the conversation. |
Get started immediately with Layercode’s optimized backend powered by Gemini Flash 2.0. Our hosted backend provides:
- Ultra-low latency responses
- Optimized for real-time conversation
- Zero backend setup required
| Setting | Description |
|---|---|
| LLM Prompt | Configure the personality and behavior of your AI assistant. |
| Welcome Message | Configure the message your AI will speak when the conversation first starts. If disabled, the user starts the conversation. |
Integrate your own backend with a simple webhook. Learn how to connect your backend →.
Benefits of using your own backend include:
- Complete control your voice AI response
- Integrate with any backend language or framework, deployed to your own infrastructure
- Total visibility into the AI’s response generation
- Use LLM providers and agent libraries you already know and love
- Use tools, MCPs and data stores you already have
Testing Your Pipeline
Click the “Try it out” button on your pipeline to visit the Playground.
The Playground is a pre-built frontend voice UI for testing out your voice pipeline. If you decide to connect your own backend, this is a great place to test it out. Even if you build your own frontend voice UI, the Playground will still work as a direct way to test your pipeline.
Next Steps
Congratulations! You’ve created your first voice pipeline. Now you can integrate it into your application.
If you are building web or mobile based voice experience, follow our guide below. You can also choose to connect your backend to your pipeline to control the AI’s response (instead of using our hosted backend). This gives you complete control over the AI’s response and allows you to use your own LLM provider and agent libraries.