PlayHT: The Generative AI voice platform

TL;DR At PlayHT we’re building our own Large Language Speech Models that are capable of generating highly expressive and emotional speech, and companies like Amazon, RedBull, Volvo , Hyundai & SalesForce use our platform to create voice content at a fraction of the cost without compromising on quality.

Hello everyone, we’re Mahmoud and Hammad, and we’re building generative voice models that can generate true human-level speech that can be used in media and content creation.

Some of the most common use cases of PlayHT are -

1. Voice cloning - media and gaming companies use our voice cloning to clone actors/character voices and use them in post-production.
2. Voiceovers - from YouTubers to training & Elearning teams at companies use our platform for creating voiceovers for their videos, employees and customers.
3. API - developers and companies use our API to create dynamic content or add interactive voice experiences to their applications.

🎧 Listen to these audio samples generated by our models → https://youtu.be/aL_hmxTLHiM

The Problem

• Recording and updating voiceovers is time consuming and tedious work. If you’ve ever tried recording one, you know what we’re talking about.
• Hiring voice actors is expensive. They charge $100-$150 per 75-100 words, plus royalty fee.
• Updating the audio in the future is a hassle because you have to get the same voice.

Our Solution

Highly expressive generative voice AI with voice cloning capabilities.

We’re solving this problem by building voice models that are trained on LLMs so they can understand the context of the text passed to them. The models are capable of cloning any voice with near perfect results with just 10 minutes of recordings.

There are two ways we offer our models -

1. We offer a diverse library of 100+ high quality fine-tuned voices that creators can start using right away.
2. We allowing creators to clone their voices and use those for creating new or update existing content.

We’ve built a text to voice editor and an API for people to access and interact with these voices.

Our story

• We started Play.ht with a chrome extension to listen to articles.
• As a way to grow, we built audio player widgets for bloggers that had a ‘powered by play.ht’ link. This turned into a product that people started paying for.
• A big problem bloggers had was to edit the text before converting it to audio. At that time, we piggy backed on other TTS APIs and built a robust TTS editor to help solve this issue.
• People started using our editor to create voiceovers and we realized a bigger need and made a pivot to build our own TTS models that are now used by thousands of people from individual creators to teams at Fortune 500 companies.

Our Ask 🙏

1. We’ve opened up access to our voice cloning tool for the launch period. Try cloning your voice and share an audio sample on social to help spread the word.
2. If you run a popular YouTube channel, podcast, newsletter or have lots of Twitter followers - we’d love to clone your voice and show it to your audience!
3. If you or anyone you know might be interested in using our models please introduce us to them or reach out to us.
4. We offer an API that gives access to our voices, please try it out and share any feedback you have - https://playht.github.io/api-docs-generator/#ultra-realistic-voices-api
5. Interested in working with us? Reach out!

You can reach out us at hammad@play.ht and mahmoud@play.ht.