Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: Neural text to speech with dozens of celebrity voices

vocodes.com

Post reply on HN
Print view View original

141–150 of 174 posts

Re: Show HN: Neural text to speech with dozens of celebrity voices

#141

Post by maccormack »

echelon wrote: post #66
spdustin wrote: post #63

Earlier quoted context omitted.

> "…Instead of using graphemes, I'm using ARPABET phonemes…" Is this why some examples I tried seemed to skip some of the words?

Exactly. If you type "I am a dangerous asdhfjahdsff velociraptor, rawr." There aren't entries for - asdhfjahdsff - rawr I added around 500 new words, but I missed a lot of stuff. The ultimate fix is to have grapheme -> phoneme prediction so that all unseen words can be mapped to potential phonemes (polyphones).

I tried "Watch as the cat sniffs the flower, eats it, and then vomits. This is classic feline behavior" with Attenborough. He seems to slip into a bit of a German accent on the second sentence. What's the cause of that?

Thanks for sharing, though. Very interesting project!

Re: Show HN: Neural text to speech with dozens of celebrity voices

#142

Post by ironmagma »

SkyBelow wrote: post #87
dralley wrote: post #46

Earlier quoted context omitted.

> On a more positive side to this technology. I'm not sure that making it easier to profit off of the likeness of others is a positive side. If it's legal for indie studios to do, it's legal for 20th Century Fox, Universal, and so forth.

It reduces cost to produce a specific good. I would think it would be measured similar to other technology advancements that do the same. Good for society, bad for the craftsman that were made obsolete, overall net positive. This is purposefully not counting in the effect of being able to fake people and the damage that does to society, but I think that was implied by the previous poster specifying looking for the mo…

It’s not so obvious that it’s a net positive. Music for instance is way easier to get ahold of now, but we also don’t get to enjoy sophisticated long-form music as much, owing to how there’s no money in making it.

Re: Show HN: Neural text to speech with dozens of celebrity voices

#143

Post by searchableguy »

azinman2 wrote: post #101
searchableguy wrote: post #100

Earlier quoted context omitted.

I don't agree with the solution unless we as a society stop putting trust on any digital media but at that point, this is not necessary. Many governments would love to use this tech so they have incentive to stop others from using it and still let people believe in digital evidence by putting half assed solution like the one you proposed. The cat is out of the bag. Digital media should not be trusted blindly.

I don’t appreciate the condescension. It’s half assed in that I put a rough direction out there for conversation for a huge problem that will inevitably cause social problems. Your reply isn’t in line with HN guidelines and certainly doesn’t make me want to participate in conversation with you about important topics.

I am really sorry. I could have worded it better. I didn't meant to be condescending to you. My reaction was more to the government.

Governments have a history of putting up solutions that work in their favor by selective enforcement and securing power to self. Encryption debate is one such example.

Re: Show HN: Neural text to speech with dozens of celebrity voices

#144

Post by superasn »

Guess the server is overloaded now. All I'm seeing are errors.

Tip: It's a cool idea to put some ready made samples under the photos. A lot of people like myself only want to hear some demos and pre saved mp3 samples are more than sufficient for that sort of thing. It will also help reduce your server loads.

Re: Show HN: Neural text to speech with dozens of celebrity voices

#145

Post by superasn »

echelon wrote: post #13
Firerouge wrote: post #3

I skimmed your about, where you mention it as a hobby demo of your deep work. Do you have a GitHub or technical documentation about how you build this sort of thing to work at scale?

I can make a blog post later, but at a high level: A rust TTS server hosts two models: a mel inference model and a mel inversion model. The ones I'm using are glow-tts and melgan. They fit together back to back in a pipeline. I chose these models not for their fidelity, but for their performance. They're 10x faster at inference than Tacotron 2. If you want something that sounds amazing, you're better off with a dense…

> I can make a blog post later, but at a high level:

A detailed blog post about this would be amazing! I wish there was a hn bot like Reddit bots to ping me when you do post it so i don't miss it.

Re: Show HN: Neural text to speech with dozens of celebrity voices

#150

Post by chrispauley »

echelon wrote: post #8

This is my pandemic side project, and I'll be happy to answer any questions about it.

Extremely fun little tool! Anywhere I can read about the techniques involve?

An interesting quirk: some words seem to get dropped entirely? for example the word "cleverer" and any word with a hyphen.

Really fun to start with a quote from one person and switch between voices to hear others recite the same line. Alan Rickman doing lines from Aladdin as Iago is pretty funny

Post reply on HN