Subscribe
Sign in
PlayHT-Turbo — The fastest generative AI Text-to-Speech API
PlayHT-Turbo, a new blazing-fast Conversational AI Text-to-Speech model with <300ms latency! Supports Input text streaming from LLMs, Output audio streaming, Cloning of any voice & accent.
20
Replies
Irakli Vekua
I tried it out and I am impressed. Especially, by the voice clone. It just works. Kudos!
Share
@sentry_co it's actually pretty good. I first used PlayHT around the middle of the year (this should be an even better and more powerful version), but even that version was scary accurate. I trained it on my voice (from podcast episodes of me speaking) and it could replicate my voice down to the pacing and pronunciation of certain words - if I didn't tell anyone it was AI, they would think it was me speaking (that was scary)
The voice and accent cloning feature has immense potential, especially in creating personalized user experiences and for multilingual support. Can't wait to integrate this into our workflow and see how it enhances user engagement!
Impressive! Can't wait to see how PlayHT-Turbo can help enhance the conversational AI experience. What are its main benefits and features that set it apart from other similar models?
The generated quality is superb and has variants and emotions in it. It is 10x better than Google's TTS, but also 10x more expensive compared to Wavenet. Do you have a plan to optimize the pricing? The pricing is the main blocking point for applying massive-user products at this moment.
I have used this product for the last couple of months and I'm really impressed. I saw them improving their product all the time, their customer service is very kind and helpful! Keep going, guys! Good luck!
Hey congrats on launching on product Hunt today! Interesting technology that you have here with PlayHT!
I was playing around with the models a bit, but I'm having a few issues. Examples:
- "The statue, made from 250 tonnes of steel, 300 litres of gold paint, and 1550 cubic meters of concrete" pronounces "fifteen-fifty cubic meters..." instead of "one thousand fifty ..."
- "an 8-hour journey, culminating in a 272-step ascent " makes a very weird long pause in the "272-step" part.
a few others which I haven't documented. Any help?
Guys this is insane!! I've been looking for weeks to a proper Text-to-Speech technology to create ultra-realistic voices.
Have been playing around now with your studio and - at least with English - it sounds insanely good!
I'll try out also non-english voices and then have a look at the pricing. I'm totally considering you guys for our Touring app!
I've already used PlayHT for a couple months - their previous version was so good at replicating human voices it's almost scary. Fed it some of my podcast audios, and it could speak (very closely) like me - down to the laughing and pacing.
Congrats on the launch of a more powerful version - gotta try it out myself
Playht Turbo seems like a game-changer. How does it adapt to various industries' needs, from e-learning to entertainment?
