sipjca

I agree, but what about the training data that goes into it (intentional poisoning of the training data, for a variety of reasons, $, power, etc.)

in some sense i think it shows having taste. sometimes taste can be dangerous. you can only have a taste for say rock climbing equipment, by rock climbing, which exposes you to danger.

looks like a paid and closed source fork of the free and open source project Handy: https://github.com/cjpais/Handy

can't say for sure, but a lot of the UI (and text) is quite familiar. the history page is a near rip off which is a giveaway.

i believe the mit license should be distributed since it's almost certainly a derivative work.

"The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software."

I can't confirm if the license is infact distributed, since I would have to pay $50, which quite frankly I'm not going to do.

a bit sad to see a ui reskin claimed as original work. the reskin is totally fine, but I believe the license must be distributed. i believe in the proliferation of this software so im happy to see this overall (it's good enough someone wants to charge for it! that's a big win!) but it's just a bit of a shame how this project has gone about it imo.

seems like a flawed assumption when the cost of tokens -> 0

that would be certifiably insane

Maybe this is true? But it's not clear to me this methodology will ever be quite as good as native tool calling. Or maybe I don't know the benchmark well enough, I just assume it's vision based

Perhaps Tesla FSD is a similar example where in practice self driving with vision should be possible (humans), but is fundamentally harder and more error prone than having better data. It seems to me very error prone and expensive in tokens to use computer screens as a fundamental unit.

But at the same rate, I'm sure there are many tasks which could be automated as well, so shrug

It also uses one of the fastest and most accurate on the ASR leaderboard, Parakeet.

Fair enough, most of the critical code in this case is written in Rust. A Rust transcription library popped out of the project `transcription-rs`. And there is a real-time audio library I'd like to put out which allows for filters. I could have called out to ffmpeg or similar, but I chose to implement an audio pipeline myself (for better or worse)

So makes sense, but there are benefits to writing a desktop application backend in Rust for the ecosystem as well.

Oh interesting. I totally misread the original comment, I didn't realize you're talking about RAM usage. 120MB is quite a lot. This surprises me too. There's nothing fancy going on really until the model is chosen.

Curious your use case, I now have quite a lot of experience with releasing desktop apps, and I have done some accessibility work as well, and may be curious to put together a TTS toolkit as well into a desktop app (or even Handy)

You're awesome Vlad!

Right now there is fairly minimal processing done to the audio. There is a VAD filter to reduce the non-speech areas. But there is no noise-reduction as such. The audio pipeline could support it though, so if you know any good real time noise reduction filters let me know. Would love to improve the SNR into the models

It's primarily this. I'm a novice Rust developer and really would like to improve the code quality across the board, and some of this comes to attracting the right kind of developers to help. Maybe "Rust" in the title helps, maybe it doesn't. Clearly HN doesn't like it and that's okay.

I stated my need for help on the about page as well

> This is my first Rust project, and it shows. There are bugs, rough edges, and architectural decisions that could be better. I’m documenting the known issues openly because I want everyone to understand what they are getting into, and encourage improvement in the project.

What computer are you using? You really should give Parakeet a try, I find it runs in a few hundred milliseconds even on a Skylake i5 from 10 years ago.

thank you!!!

To be fair on the other side there is a fair lack of specific ML inference libraries in Rust, and this project is pushing some of that forward with Parakeet at the very least. The Rust library `transcribe-rs` came from it and hopefully will support more models in the future.

While certainly it's not an ML project in the sense of I am not training models, the inference stack is just as important. The fact is the application does do inference using ONNX and Whisper.cpp.

Benchmarks definitely say otherwise, but my anecdotal experience says medium is the best for this application with my voice and microphone

Yes! Turn off “push to talk”, it will activate when you click the shortcut and stop when you click it again

Whisper uses Vulkan and Metal acceleration with whisper.cpp

Parakeet is currently CPU only

Next version will have it! The main branch already has I’ve just not released the next version yet

A lot of the bloat comes from dependencies like ONNX or whisper.cpp to accelerate running the model itself

While the UI is doing “nothing” most of the bloat is not from the UI

It's typescript because it is a Tauri app which uses the system webview to render the UI.

Most of the audio code/inference code is Rust or bindings to libraries like whisper.cpp

Author here!

The default recommendation is Parakeet (mainly because it runs fast on a lot more hardware), but definitely think people should experiment with different models and see what is best for them. Personally I found Whisper Medium to be far better than Turbo and Large for my speech, and Parakeet is about on par with Medium, but each have their own quirks.

I'll update the site soon!

If you prefer a more stripped down version: the original releases (0.1.0 and 0.1.1) shipped with Whisper tiny included and no auto-update feature

Hey author of Handy here! Would absolutely love any help, please let me know if there's any way I can make contributing easier!

This is super rad. Love it being Open Source, and with the option to choose local models. You’re awesome, thanks!

but what if you don't need to run it in the cloud

i mean you're also throwing away security fixed running an OS that out of date, but the person doing this probably doesn't care about security anyway.

i mean there are also lots of browser options to be fair.

should they be locking safari to the OS, definitely not. but users can just go download another browser if they are actually concerned.

How does Turing completeness feel like a lie?

Planned obsolescence is some of it, some of it is abstractions making it easier for more people to make software (at the cost of using significantly more compute) and Moore’s law being able to support those abstraction layers. Just imagine if every piece of software had to be written in C, the world would look a whole lot different.

I also think we’ve gone a bit too far into abstraction land, but hey, that’s where we are and it’s unlikely we are going back.

Turing completeness is almost an unrelated concept in all of this if you ask me, and if anything it’s because of completeness that has driven higher and higher memory and compute requirements.