Podcast recording and editing platform Podcastle is making a significant move into the AI-powered text-to-speech market with the release of its own AI model, Asyncflow v1.0. This new technology allows for the conversion of text into voice clips narrated by AI, with an impressive offering of over 450 AI voices. Additionally, an API for developers will be available, enabling them to directly integrate the text-to-speech model into their applications.

The startup's founder, Arto Yeritsyan, revealed that the company had always aimed to develop a text-to-speech model but was hindered by the high costs of training and data requirements. However, with recent advancements in large language models, Podcastle was able to achieve a breakthrough last year, making it possible to build a high-quality voice model without requiring extensive data.

The release of Asyncflow v1.0 puts Podcastle in the same league as other startups, including ElevenLabs, Speechify, and WellSaid, which have also developed AI-powered text-to-speech technology. This technology has far-reaching implications, spanning various use cases such as marketing, advertisement, content creation, education, and corporate training.

In terms of pricing, Podcastle charges around $40 per 500 minutes of text-to-speech conversion, significantly lower than ElevenLabs' $99 for the same service. Yeritsyan attributes this competitive pricing to the company's ability to keep training and inference costs low, thanks to its innovative approach to developing the technology.

Podcastle's voice cloning feature is also receiving an upgrade, allowing for a faster process of training. Previously, the training process required reading around 70 different sentences, but now it only needs a few seconds of recording to create a clone of one's voice. This new process utilizes Podcastle's Magic Dust AI, released last year, to improve audio recording quality.

In testing, the voice created with the new process sounded slightly robotic but managed to mimic the tone. Podcastle assures that it will continue to improve this feature over time, allowing users to train different samples of their voice to achieve varying results.

According to Yeritsyan, having tools for audio, video, podcasts, and AI-powered narration under one redesigned site will give Podcastle an edge over its competitors. While the majority of users currently utilize Podcastle for audio content, video is rapidly gaining traction, and the company is well-positioned to capitalize on this trend.

The release of Asyncflow v1.0 marks a significant milestone for Podcastle, and its implications will be closely watched in the AI-powered text-to-speech space. As the technology continues to evolve, it will be interesting to see how Podcastle and its competitors adapt and innovate to meet the growing demands of this rapidly expanding market.