How is this an Alexa knockoff when it doesn't include any home automation features?
I made a simple Alexa knockoff that doesn’t spy on me
11–19 of 19 posts
Re: I made a simple Alexa knockoff that doesn’t spy on me
#12Good lord. The "Alexa" knockoff has no voice recognition! Based on this title style I built a google and gmail knockoff - using index cards, ink and paper and postage stamps. It doesn't spy on me - but also doesn't work at all like google or gmail! Better title: I built an LCD display that displays information?
Yeah, I was expecting to see a blog post about raspberry pi speech recognition. I was hoping it wasn't just another "look, Ma, I did speech recognition! (By sending everything to )" but I was prepared to be disappointed on that front because such posts are endemic these days. Turns out, it didn't engage with speech recognition at all, which was even more jarring.
Speaking of rpi offline speech recognition, is Jasper still the go-to? If you're willing to spring for one of the shiny new nvidia SBCs, is there an open source option that can take advantage of their power (with lower effort than "glue everything together manually with tensorflow")?
Re: I made a simple Alexa knockoff that doesn’t spy on me
#13Earlier quoted context omitted.
> voice recognition and voice commands Nor are they that hard to do with the right libraries
Unless things have changed a lot in the past few months, they actually are pretty limited locally at this point. Certainly you can do them, but you can't them particularly well.
Things are always changing and I'd be curious what you're basing that claim on.
wav2letter can transcribe locally and well.
Re: I made a simple Alexa knockoff that doesn’t spy on me
#14If this was to be called an "Alexa alternative", there should at least be Home Assistant running on that Raspberry Pi. Despite also not having voice recognition, it could then at least be able to, you know, DO something.
Re: I made a simple Alexa knockoff that doesn’t spy on me
#15Earlier quoted context omitted.
Unless things have changed a lot in the past few months, they actually are pretty limited locally at this point. Certainly you can do them, but you can't them particularly well.
Things are always changing and I'd be curious what you're basing that claim on. wav2letter can transcribe locally and well.
Mostly from playing a bit with Sphinx and loosely following the space. But as you say things are always changing.
I will note that different people have different definitions of "good" in this space. For example, I've used online services like otter.ai and, while they're useful for certain purposes, they're still much worse than a human for transcribing. They're certainly not worth my time to clean them up to publish an actual transcript.
Re: I made a simple Alexa knockoff that doesn’t spy on me
#16It's a knockoff, but doesn't include NLU or ASR capabilities so really it's a shitty tablet on a kickstand I could have purchased for roughly the same price with more features.
Re: I made a simple Alexa knockoff that doesn’t spy on me
#17Good lord. The "Alexa" knockoff has no voice recognition! Based on this title style I built a google and gmail knockoff - using index cards, ink and paper and postage stamps. It doesn't spy on me - but also doesn't work at all like google or gmail! Better title: I built an LCD display that displays information?
Yeah, I was expecting to see a blog post about raspberry pi speech recognition. I was hoping it wasn't just another "look, Ma, I did speech recognition! (By sending everything to )" but I was prepared to be disappointed on that front because such posts are endemic these days. Turns out, it didn't engage with speech recognition at all, which was even more jarring. Speaking of rpi offline speech recognition, is Jasper…
Jasper is very old and not very accurate. For offline recognition on RPi try something like Vosk https://github.com/alphacep/vosk-api
Re: I made a simple Alexa knockoff that doesn’t spy on me
#18Earlier quoted context omitted.
Things are always changing and I'd be curious what you're basing that claim on. wav2letter can transcribe locally and well.
Mostly from playing a bit with Sphinx and loosely following the space. But as you say things are always changing. I will note that different people have different definitions of "good" in this space. For example, I've used online services like otter.ai and, while they're useful for certain purposes, they're still much worse than a human for transcribing. They're certainly not worth my time to clean them up to publish…
but for something like Alexa which is designed to recognize specific commands it would be easy to design something with relatively good accuracy
I have not tried the other [otter.ai] service but even with speech to text on my iPhone, which to note is not running luckily [locally], it seems pretty accurate at transcribing what i say
^ the above was transcribed on an iPhone and the errors are bracketed.
Re: I made a simple Alexa knockoff that doesn’t spy on me
#19Earlier quoted context omitted.
Yeah, I was expecting to see a blog post about raspberry pi speech recognition. I was hoping it wasn't just another "look, Ma, I did speech recognition! (By sending everything to )" but I was prepared to be disappointed on that front because such posts are endemic these days. Turns out, it didn't engage with speech recognition at all, which was even more jarring. Speaking of rpi offline speech recognition, is Jasper…
Jasper is very old and not very accurate. For offline recognition on RPi try something like Vosk https://github.com/alphacep/vosk-api
Thanks, that's exactly the kind of tip I was looking for!