Live data from Hacker News

Show HN: Three new Kitten TTS models – smallest less than 25MB

github.com

161–170 of 201 posts

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#161
Very cool :) Look forward to trying it out

Maybe a dumb and slightly tangential question, (I don't mean this as a criticism!) but why not release a command line executable?

Even the API looks like what you'd see in a manpage.

I get it wouldn't be too much work for a user to actually make something like that, I'm just curious what the thought process is

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#163
post #161

Very cool :) Look forward to trying it out Maybe a dumb and slightly tangential question, (I don't mean this as a criticism!) but why not release a command line executable? Even the API looks like what you'd see in a manpage. I get it wouldn't be too much work for a user to actually make something like that, I'm just curious what the thought process is

great idea, we'll do that too. we just decided to launch an onnx first and get some feedback. we'll be simplifying the process of running it everywhere including a command line executable.

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#164
post #4

Is it English only?

as of now its english only. the training for multilingual model is underway and should be out in April! what languages are you most interested in? Right now, we are providing deployments for custom languages + voices through support form on the github.

Here to suggest the Bengali language! It has the 7th largest speaker base worldwide but often ignored by tech companies, sadly.

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#165

Earlier quoted context omitted.

Nowadays these frustrations shouldn't be a thing any more. If the author used uv, the script would be able to install its own dependencies and just work.

yeah let me add uv and conda support to make it easier.

One of the most responsive developers I’ve ever seen, kudos

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#168
post #154

Is this open-source or open-weights ML?

yes, indeed. we are working on adding mit licensed phonemizers too by this weekend, so you'll be able to use these models as you like :)

I think you misunderstood the question. I guess its only open-weights not open-source then.

For some insight into the original question, take a look at the Debian ML policy:

https://salsa.debian.org/deeplearning-team/ml-policy

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#169
The size/quality tradeoff here is interesting. 25MB for a TTS model that's usable is a real achievement, but the practical bottleneck for most edge deployments isn't model size -- it's the inference latency on low-power hardware and the audio streaming architecture around it. Curious how this performs on something like a Raspberry Pi 4 for real-time synthesis. The voice quality tradeoff at that size usually shows up most in prosody and sentence-final intonation rather than phoneme accuracy.

Re: Show HN: Three new Kitten TTS models – smallest less than 25MB

#170

Only American voices? For some reason I'm only interested in Irish, British or Welsh accents. American is a no

minor nit to pick: Welsh accents are British accents as Wales is in Britain. In fact by some definitions it's the most British part.

People from outside the UK often use British as synonymous with English, and in the context of accents, often a South East English accent or some sort of Received Pronunciation (RP) accent. Technically a "British" accent could be from anywhere in England, Scotland, or Wales, and therefore by extension might not even be the English language.

While I'm here, since it's generally confusing, the UK is Great Britain and Northern Ireland. Great Britain is England, Scotland, and Wales.

Post reply on HN