Live data from Hacker News

Qwen3-TTS family is now open sourced: Voice design, clone, and generation

qwen.ai

201–210 of 229 posts

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#201

Earlier quoted context omitted.

I just followed the Quickstart[1] in the GitHub repo, refreshingly straight forward. Using the pip package worked fine, as did installing the editable version using the git repository. Just install the CUDA version of PyTorch[2] first. The HF demo is very similar to the GitHub demo, so easy to try out. pip install torch torchvision --index-url https://download.pytorch.org/whl/cu128 pip install qwen3-tts qwen-tts-demo…

It flat didn't work for me on mps. CUDA only until someone patches it.

Demo ran fine, if very slowly, with CPU-only using "--device cpu" for me. It defaults to CUDA though.

Try using mps I guess, I saw multiple references to code checking if device is not mps, so seems like it should be supported. If not, CPU.

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#202
Curious how it compares to last week’s release of Kyutai’s Pocket-TTS [1] which is just 100M params, and excellent in both speed and quality (English only). I use it in my voice plugin [2] for quick voice updates in Claude Code.

[1] https://github.com/kyutai-labs/pocket-tts

[2] https://github.com/pchalasani/claude-code-tools?tab=readme-o...

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#203

Earlier quoted context omitted.

> This feels like one of those tropes that keeps showing up whenever new tech comes out. And this itself is another tired trope. Just because you can pattern match and observe that things repeatedly went a certain way in the past, doesn't mean that all future applications of said pattern will play out the same way. On occasion entire industries have been obliterated without a trace by technological advancement. We ca…

Give me one aspect in which that has actually happened? I'm wracking my brains but can't think of one. We are a weird species in that even if we could replace ourselves our fascination with ourselves means that we don't ever do it. Cars and bicycles have replaced our ability to travel at great and small distances and yet we still have track events culminating in the olympics.

Sure, things continue to persist as a hobby, a curiosity, a bespoke luxury, or the like. But that's not at all the same thing as an industry. Only the latter is relevant if we're talking about the economy and employment prospects and making a living and such.

It's a bit tricky to come up with concrete examples on the spot, in particular because drawing a line around a given industry or type of work is largely subjective. I could point to blacksmithing and someone could object that we still have metalworkers. But we don't have individual craftsmen hammering out pieces anymore. Someone might still object that an individual babysitting a CNC machine is analogous but somehow it feels materially different to me.

Leather workers are another likely example. To my mind that's materially different from a seamstress, a job that itself has had large parts of the tasks automated.

Horses might be a good example. Buggies and carriages replaced by the engine. Most of the transportation counterparts still exist but I don't think mechanics are really a valid counterpart to horse tenders and all the (historic) economic activity associated with that. Sure a few rich people keep race horses but that's the sort of luxury I was referring to above. The number of related job positions is a tiny fraction of what it was historically and exists almost solely for the purpose of entertaining rich people.

Historically the skill floor only crept up at a fairly slow rate so the vast majority of those displaced found new sectors to work in. But the rate of increase appears to have picked up to an almost unbelievable clip (we're literally in the midst of redefining the roles of software developers of all things, one of the highest skilled "bulk" jobs out there). It should be obvious that if things keep up the way they've been going then we're going to hit a ceiling for humans as a species not so long from now.

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#204
post #192

I can't quite figure this out: Can you save a generated voice for reuse later? The mlx-audio I looked at seems to take the text itself in every interface and doesn't expose it as a separate object. (I can dive deeper, but wanted to check if anyone's done it already)

You could pipe the output to an audio file with ffmpeg or pyaudio and save it locally

I don't want to save the audio. I want to save the voice model so I can use it for many different texts, for consistency.

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#205

Earlier quoted context omitted.

Pocket TTS is much smaller: 100M parameters versus 600–1800M.

Ah right so I guess qwen3-tts isn't going to work for cpu-only like pocket TTS can(?)

The current code doesn't appear very optimized. Running on CPU-only it only uses four threads for example, nowhere close to saturating all my cores.

As a result it's dog slow on CPU only, like 3-4 minutes to produce a 3 second clip, and still significantly less than real-time on my 5090 using only 30% of the GPU.

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#206

Earlier quoted context omitted.

Thanks. All modern TTS for Japanese are useless failures.

Just use that green guy, classical TTS ain't broke nanoda.

I’m building Manabi Reader and would need to contact them for an enterprise contract. Annoying licensing but maybe inevitable

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#207

Earlier quoted context omitted.

What happens to lyricless electronica if suddenly every electronic artist has quality vocal-backing? Oh no . Maybe we did frig this up.

On the other hand, maybe we'll get models capable of removing the lyrics from things without damaging the rest of the audio. Or better yet, replacing the lyrics with a new instrument. So it might yet work out in our favor.

This was one of the first things they were doing with neural nets,

and there are even a couple SaaS options for it now.

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#208

Earlier quoted context omitted.

What happens to lyricless electronica if suddenly every electronic artist has quality vocal-backing? Oh no . Maybe we did frig this up.

More choices for artists is not a bad thing.

Indeed.

But it does change who can be an artist in each niche,

and that’s been interesting to briefly pause and consider here with the community.

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#209

Earlier quoted context omitted.

This is terrifying. With this and z-image-turbo, we've crossed a chasm. And a very deep one. We are currently protected by screens, we can, and should assume everything behind a screen is fake unless rigorously (and systematically, i.e. cryptographically) proven otherwise. We're sleepwalking into this, not enough people know about it.

> This is terrifying. Far more terrifying is Big Tech having access to a closed version of the same models, in the hands of powerful people with a history of unethical behavior (i.e. Zuckerberg's "Dumb Fucks" comments). In fact it's a miracle and a bit ironic that the Chinese would be the ones to release a plethora of capable open source models, instead of the scraps like we've seen from Google, Meta, OpenAI, etc.

>Far more terrifying is Big Tech having access to a closed version of the same models, in the hands of powerful people with a history of unethical behavior (i.e. Zuckerberg's "Dumb Fucks" comments).

Lol what exactly do you think Zuck would do with your voice, drain your bank account??

Re: Qwen3-TTS family is now open sourced: Voice design, clone, and generation

#210
post #129

Earlier quoted context omitted.

Your GitHub profile is... disturbing. 1,354 commits and 464 pull requests in January so far. Regardless of how productive those numbers may seem, that amount of code being published so quickly is concerning, to say the least. It couldn't have possibly been reviewed by a human or properly tested. If this is the future of software development, society is cooked.

It's mostly trying out my orchestration system ( https://github.com/mohsen1/claude-code-orchestrator and https://github.com/mohsen1/claude-orchestrator-action ) in a repo using GH_PAT. Stuff like this: https://github.com/mohsen1/claude-code-orchestrator-e2e-test... Yes, the idea is to really, fully automate software engineering. I don't know if I am going to be successful but I'm on vacation and having fun! if Opus 4…

On the contrary, that actually is pretty cool. z.ai subscription is cheap enough that I'm thinking to run it 24/7 too. Curious if you've tried any other AI orchestration tools like Gas Town? What made you decide to build your own, and how is it working for you so far?
Post reply on HN