Podcast recording and processing platform Podcastle now comes together with other corporations within the AI-driven text-to-language race by calling its own AI model Asyncflow V1.0. An API for developers can be available so that you could integrate the text-to-language model directly into your apps.
Thanks to the brand new model, the corporate can offer greater than 450 AI voices that may tell their text. The startup said that it developed the technology and the model in order that its training and inference costs are low, which provides it a bonus over the competitors.
With the move, Podcastle connects quite a few startups, including Elevenlabs, Speechify and Wellsaid, which have developed technology and AI models with a view to convert any type of text right into a language clip told by AI. This technology includes applications equivalent to marketing, promoting, content acquisition, education and company training.
The founding father of Podcastle, Arto Yeritsyan, told Techcrunch that the corporate has all the time desired to create a text-to-language model, but the prices for training and data requirements for this were very high.
“We wanted to construct a strong model for Speech model since our foundation. However, the event costs were very high. Thanks to the recent developments in the nice -language model, we were capable of make a breakthrough last yr to get to a spot where we could create a high -quality voice model without having plenty of data, ”said Yeritsyan.
The company was also supported in its efforts by its Serie A donation share of 13.5 million US dollars last yr.
Yeritsyan said that podcastle calculates around $ 40 per 500 minutes of text-to-language conversion, but Elflabs calculates $ 99 for a similar.
Podcastle's Voice Cloning function also gets an upgrade to create a faster process for training.
Previously, the training process was reading around 70 different sentences. Now you simply need a couple of seconds of recordings from you to create a clone of your voice. The latest process also used Podcastle's Magic Dust Ai, which was published last yr to enhance the standard of the audio recording.
In our tests, the voice generated with the brand new processes sounded a bit robot, regardless that it imitated our tone. The company said that it’s going to improve the function over time. You may train different samples of your voice to realize different results.
Podcastle said that aside from the prices, tools for audio, video, podcasts and AI-driven narrative under a redesigned website, there can be a bonus over the competitors. Yeritsyan said that nearly all of users use podcastle to work on audio content, but additionally meet up with the video.
