New in Resemble: GPT-3 with Synthetic Voices

Mar 30, 2021

Over the last few months, we’ve been blown away by the unique use-cases that are being built on Resemble AI. Last year, we wrote about how we imagined GPT-3 and custom synthetic voices would work together to bring artificial characters to life.

Today, we’re excited to announce that we have native support for GPT-3 right within our powerful content editor.

 

The Basics

Getting started with our GPT-3 and AI Voice integration in the editor is super simple. You must write a minimum of two blocks of content for the AI to know the context of what it should generate next.

 

GPT-3 and Synthetic Voices

 

When you hover over the right toolbar on the block that you’re in, you’ll notice a pencil with “AI-Generated Autocompletion” on top. Simply click the button to let the AI generate the next bit of text. Within a second, you’ll see the AI write out the text in the block you’re in:

GPT-3 and Voice Cloning

 

Now, you’re not always building a Helpdesk or a contact center, so we’ve designed built-in categories that you can use to generate content for different use-cases. On the right sidebar, you’ll see three categories to choose from:

GPT-3 Uses

Conversational Dialogue

The conversational dialogue category creates natural text that flows within a dialogue. Whether you’re making a game or an Alexa Skill, the AI will craft dialogues that proceed naturally. Once the text is generated, simply click on the Play button to listen to your content with the AI voices.

GPT-3 Conversation

Some tips to consider:

  • Write a few passages before you get the AI to generate content.
  • It’s good practice to describe each character first with their voices. This gives the AI some idea about the persona it is generating content for.
  • Try to start the dialogue with a few words before generating the content.
  • Make sure you review every passage to ensure that the quality is high and is safe for the experience that you are creating.

Advertising

Crafting the perfect ad copy used to be difficult and laborious. Not anymore. GPT-3’s text AI engine will automatically craft copy that fits your product name and description. The more description you provide for your product, the better the AI will be at predicting what it thinks are key attributes for advertising copy. You can always make any edits you want in the text box to make it sound even better.

What’s new is that now you can even choose from our wide selection of professional voices to make your ads stand out. If you have a voiceover artist you already work with, you can create their synthetic voice and vocalize the text perfectly without multiple takes in the studio.

GPT-3 Ads

Some tips to craft the best advertising copy:

  • Make sure that the description of your product includes key elements that you want the AI to touch on.
  • Generate many utterances until you find the copy that you like the most.
  • Combine ideas that the AI generates to craft your own advertising copy.

Helpdesk

Using AI to generate copy for your contact center or IVR flow has never been easier. We’ve crafted the helpdesk category to produce content that demonstrates empathy and helpfulness. By providing it a few utterances from your own corpus, the AI will generate meaningful content that will fit right into your dialog engine.

You can then choose from our selection of friendly voices to give your customers the best experience that’s tailored to their exact needs.

GPT-3 powered Helpdesk

Some helpful tips to design your helpdesk:

  • Think of the tone of your voice with the content that you insert before you generate using the AI.
  • Describe your service in the blocks that come before the AI-generated text.

Risks and Rules

Similar to AI voices, we take AI ethics very seriously. That’s why when we partnered up with OpenAI, we built a content filter that prevents users from generating text that violates our terms of service. We will automatically detect content that is political, religious, or talking about a protected class such as race or nationality. We also prevent text that contains profane language, prejudiced or hateful language, something that could be NSFW, or text that portrays certain groups/people in a harmful manner. This will make sure that the text that’s generated is acceptable for social standards.

We can’t wait to see what you’ll build with AI when you combine text with voices powered by Resemble and GPT-3.

More From This Category

Our Commitment to Consent

Our Commitment to Consent

Remember when creating a synthetic voice meant hours in a studio, carefully recording every syllable? Now, with a few clicks, you can clone anyone's voice. It's mind-blowing tech. But with great power comes great responsibility. At Resemble, we've always believed that...

read more
Introducing ‘Edit’ by Resemble AI: Say No More Beeps

Introducing ‘Edit’ by Resemble AI: Say No More Beeps

In audio production, mistakes are inevitable. You’ve wrapped up a recording session, but then you notice a mispronounced word, an awkward pause, or a phrase that just doesn’t flow right. The frustration kicks in—do you re-record the whole segment, or do you spend...

read more
Introducing Resemble Identity & Audio Intelligence

Introducing Resemble Identity & Audio Intelligence

We're excited to unveil two groundbreaking models designed to revolutionize your interaction with audio: Resemble Identity and Resemble Audio Intelligence. These tools enhance speaker recognition, real-time analysis, voice-based authentication, and more. Resemble...

read more