Voximplant Platform rolls out a beta version of Avatar, an out-of-the-box NLP for automated omnichannel communications that actually works.

We’re thrilled to announce the launch of the Avatar beta version. The new product offers developers NLP capabilities natively integrated into Voximplant Platform.

Focus on Business Logic only using an All-in-One Platform 

With Voximplant Avatar, you can easily integrate NLP functionality and get to market faster and cheaper. There’s no need to deal with multiple vendors, write complex backend logic, or set up integrations and data management workflows. 

Better yet, there are no personal data transfers between vendors and syncing their data policies – this is especially important in the age of GDPR.

We offer a complete and flexible communications stack all in one place: 

  • Take telephony to handle phone calls
  • Add speech synthesis and recognition to tie audio streams to dialog management system
  • Set up webhooks to integrate your CRM and connect customer data
  • Lastly, embed an ML engine to power your voicebot

Focus on your customer experience, we take care of the heavy lifting so you don’t have to.  

Jump into State-of-the-Art Machine Learning Models

Many NLP providers offer outdated ML models and scenarios. There are cases when voicebots lock callers in the intent block. 

For instance, imagine a banking customer calling customer services to transfer funds from one account to another. During the conversation, the customer wants to check their account balance and ensure there are enough funds to make the transfer. An outdated voicebot won’t be able to provide account information so the customer’s problem will remain unresolved.

We enable you to create more human-like interactions for your customers. If a caller wants to move to another conversation path, Avatar will easily handle it. 

Create Complex Communications Flows with JS

Previously, you had to be an expert in telecommunication technologies, software engineering, and machine learning to build a voicebot. Voximplant Avatar makes this easy and can be used by almost any developer who knows basic JavaScript.

How it works:

  1. Intent classification - Avatar categorizes all the heard phrases according to the prescribed intents.
  2. Information extraction - Avatar extracts structured information such as numbers and dates from the speech and return it in a machine-readable format.
  3. Scenario processing - The information goes to the scenario where Avatat determines how to answer the question and whether to make a request to the CRM.

Debug and Improve Your Neural Network in a Text Mode

Machine learning is not always perfect. Sometimes it makes mistakes, but you can easily fine-tune your model using real data from the conversation history. 


With small adjustments, you can easily train, evolve and expand models over time. We use a state-of-the-art neural network to power our ML engine, fine-tuned on conversational data to have the best-in-class level of quality.

Create a Voicebot in Five Steps

  1. Login to the platform account. If you don’t have one, you can sign up here.
  2. Create an empty Avatar. Go to the Avatars section, create a bot and give it a name. After that, you’ll see the empty Avatar with predefined basic intents and a basic dialog scenario.
  3. Add scenario-specific intents. Add examples of utterances users can express this intention. You can also add default responses for these requests.
  4. Create a dialog scenario. We’ll give you the default conversation scenario, which you can customize for your use case.
  5. Integrate into a platform scenario. Customize your integration or use a pre-made solution.

That’s it! You can find the detailed guide in our documentation.

Leave Beta Testing Feedback

By beta testing, you’ll become an important part of the Avatars’ development. Your participation and feedback will help us release a better version of the product.

Test the new functionality and leave your feedback in our Discord.