Podcastle Unveils Text-to-Speech with 450+ AI Voices

Podcastle, the popular platform for podcast recording and editing, is jumping into the AI text-to-speech game with its new model, Asyncflow v1.0. They're not just stopping there—they're also rolling out an API for developers, making it a breeze to plug this tech into their own apps.

With Asyncflow v1.0, Podcastle's now boasting over 450 AI voices ready to read your text out loud. The cool part? They've managed to keep the training and inference costs down, which gives them a leg up on the competition.

Podcastle is now in the same league as other startups like ElevenLabs, Speechify, and WellSaid, all of which are using AI to turn text into spoken words. This tech's got applications all over the place, from marketing and ads to making content, teaching, and even training folks at work.

Arto Yeritsyan, the brains behind Podcastle, told TechCrunch that they've been wanting to dive into text-to-speech for a while now. The problem? It was just too pricey and data-hungry. But thanks to some recent breakthroughs in large language models, they finally cracked the code last year. "We've been aiming to build a solid text-to-speech model right from the start," Yeritsyan said. "But the costs were sky-high. With the latest tech, we found a way to create a top-notch voice model without needing a mountain of data."

Their $13.5 million Series A fundraise last year also helped fuel this project.

When it comes to pricing, Podcastle's pretty competitive. They're charging around $40 for 500 minutes of text-to-speech conversion, while ElevenLabs asks for $99 for the same amount.

Podcastle's voice cloning feature is getting a facelift too, making the training process a whole lot quicker. Before, you had to read out about 70 sentences to train the system. Now, it only takes a few seconds of your voice. They're using their Magic Dust AI, which they launched last year, to boost the audio quality during this new process.

When we tried it out, the voice sounded a bit robotic, but it did catch our tone pretty well. Podcastle says they're working on making it even better. You can also train the system with different voice samples to get varied results.

Podcastle reckons that, aside from keeping costs low, having everything from audio and video tools to podcasts and AI narration in one place will set them apart from the competition. Yeritsyan mentioned that while most users stick to audio content, video's starting to gain ground too.