Looking for test users for AI-Powered Voice Assistant that empowers computer control for the visually impaired

Hi,

I just signed up for it. I cannot wait to test it out. question: how does this compare to existing features such as siri and voice control? and can you check for updates in future as soon as it's released?

cool! cannot wait to test it out. Is it always going to be free? and will there be a check for updates option in the help menu?

Nice! and the subscription is optional, right? also, are there sound notifications to let you know it herd your commands just like siri?

You mentioned that it can be used to press buttons, can it also be used for clicking on links such as opening a link to a zoom meeting? say, click on OK, when the "this meeting is being recorded" dialog box pops up? another example, lets say that you are composing a message in the messages app, can you say "click on record audio," and it clicks on the button, and when you're done recording your audio message, instead of hunting for the stop button, can you say, "click stop" and it'll stop recording? if yes, I could use it in situations like this.

I forgot to ask, I know I have signed up to get access to the app, but is there a website to download the app directly?

Yes I did. I applyed on website.

Is there a youtube channel with demos? I would love to hear demos if any.

No problem. Cannot wait to test app. also, is there help available if you ever get stuck?

I assume you already know that voiceover exists?

At the moment it sounds like you're trying to create a screen reader but through voice.

I'd recommend also adding a text to nexy option that way if you don't want to talk, or can't for some reason, you could stil use all the functions.

Hi. I’ve signed up, but unfortunately don’t have a Mac. Could I still be added to the list so that I can test if/when a Windows version comes out? Also, if I do make it into the first 10 users and as you said that the subscription will be free forever for those users would I be able to get it for free on windows? If not, that’s perfectly alright.

does it have windows pc version?

To test it out if the dev has any plans of coming up with a windows version...

Does it support Intel Macs? And do I need to "guide" it through each step - example Press send on the mail interface to send the message or can I just say something like "Send an email to x with subject test and write this is a test?"

It's not for me but others seam to enjoy it. You can get a month free trial I believe: https://www.guideinteraction.com/

Yes you can read about Guide and discuss it on our Windows forum here:
https://www.applevis.com/forum/windows/guide-ai-assistant-people-who-are-blind-or-low-vision

Hi @Seregawpn. Can you comment at all about privacy and security? With an app like this you are giving over a lot of control and access to your Mac, so trust in the privacy and security is naturally very important.
Many thanks,
Dave

Hey there,
I'm hopeful I can get a slot. I signed up yesterday afternoon.
I have some really interesting and different use cases I want to test out, I won't go into detail until I have the chance too, but I feel like if successful, it could be a game changer.

I won't take the link down but will look next time for a link before posting.

Like civilization let’s say?

Just signed up. I'm an iOS and macOS developer as well and work with AI so hopefully can provide some useful feedback and suggestions.

the OP hasn't responded in a couple days, I've not put my email in this website and would advise others to stay away until this person responds.

Shiny toys are cool, but when there's not responses from the devs of those shiny new toys, I get quite wary.

Thanks for letting us know.

Sorry for the reply before, it's just you seamed very quick to answer, it never crossed my mind that you had testers to see to first.

I'm looking forward to checking out a video or two.

Greetings and salutations, I signed up via the link provided, and I haven’t heard anything back. It’s probably been a couple days, so I don’t know how this is supposed to work, or how you’re in the queue to get selected. Just thought I’d put it out there.

Awesome!

This thread is giving me vibes of a thread from some time ago where its original poster was asking people to fill in a form with personal information to join some kind of group conversation on the subject of accessibility with Apple representatives. Nothing ended up coming out of it, and I suspect that the thread was actually deleted since I can't find it in my post history. The fact that the original poster here is now deflecting and deferring responses after being very quick to reply in the beginning, coupled with the fact that they haven't even addressed the privacy question raised by an earlier commenter, makes the whole thing feel rather fishy.

Also, and excuse my negativity, but I think that, if this project actually exists, in its current form it's just a gimmick to attract investment by surfing the AI hype. As a blind Mac power user I don't think there's much that an AI agent can offer me in terms of accessibility. I do not deny that navigating inaccessible content would help a lot, but being relegated to the passenger's seat when it comes to controlling my own computer is not something that I will give up on easily. There's a lot that AI can do not only for us but for humanity in general, and I do work for a company whose founders have been impressing me with lots of good ideas that I and others have been realizing into an actual product, but fortunately, agentic crap does not seem to be on their plans.

If this project really exists and pursues a serious objective, my recommendation is to focus on how to assist us doing things rather than on how to do things for us, by dropping the agentic crap and focusing into improving the situation with poorly accessible or totally inaccessible content. Even if there was a zero percent chance of the AI hallucinating, and even if it could somehow read my mind and make sense of my ambiguous prompts, I would still prefer to retain control since it's almost always more efficient.

Hi, any updates? I hope it's ready to be tested.

OK. also, let me know when the youtube video is ready. Would love to hear a demo.

Hi, No updates? what's going on.

This is funny because I actually browsed this thread earlier today and was considering commenting on it again, but then decided against it in order for my actions to not be perceived as a form of harassment.

There are some red flags here, so I strongly recommend against filling in their form with personal data and especially installing anything requiring input device access, accessibility privileges, AppleEvents privileges, or access to the camera, microphone, screen capturing, or system audio capturing facilities at the very least until they show an actual working demo.

Honestly, it seams like one of those things where a sighted person thought, let's help the poor blind peple and didn't actually ask or hire any blind peple to get this off the ground.

If a demo comes out, I'll check it out but they've been silent for a couple of weeks now.

CoreAudio has a built-in echo cancellation feature that can be enabled using the kAudioDevicePropertyVoiceActivityDetectionEnable property of an audio device object.

The following is a section of a comment in /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/System/Library/Frameworks/CoreAudio.framework/Headers/AudioHardware.h:

    @constant       kAudioDevicePropertyVoiceActivityDetectionEnable
                        A UInt32 where 0 disables voice activity detection process and non-zero enables it.
                        Voice activity detection can be used with input audio and has echo cancellation.
                        Detection works when a process mute is used, but not with hardware mute.

The above file can also be opened in an editor like TextMate from Terminal in a more portable way using xcrun as follows:

mate `xcrun --show-sdk-path`/System/Library/Frameworks/CoreAudio.framework/Headers/AudioHardware.h

Writing a fast wavelet transform based echo cancellation solution from scratch using the SIMD API from the Swift standard library if the CoreAudio option proves insufficient is not that complicated either, however I recommend against using Apple's BNNS module from the Accelerate framework since it's not really designed to take advantage of CPU cache, plus has no support for the discrete or fast wavelet transforms so you won't be saving yourselves any work. In this case you'd transform the spectrogram produced by the wavelet transforms of the input samples and a bigger chunk of output samples into a set of bezier paths, try to match those paths in both spectrograms, remove the ones that match best from the input, and reconstruct the signal using the inverse fast wavelet transform. In theory this solution provides the best results even in noisy environments.

Another option is to use the fast cosine transform, which I think is available in Accelerate.BNNS, perform the dot product between the last input chunk and a sliding window over the last few hundreds of milliseconds of output in the frequency-domain, subtract the output vector from the input multiplied by the computed dot product where the similarity is greatest, and use the inverse fast cosine transform to reconstruct the new signal. This solution provides good results in theory even with some noise, but small differences in the amplitude of each input frequency resulting from the audio signature of the speakers may result in some output bleeding back in.

Finally the lamest solution is to just perform a time-domain correlation between the input and output samples and subtract the input from the output where the similarity is highest. This is the easiest option but may not be sufficient because any difference between the output and input signal resulting from the audio signatures of the speakers will cause bleeding and this will also perform very poorly in noisy environments.

Edited to correct, clarify, and improve my suggestions.

Why would I want to use your app when I can use voiceover to do the same thing?

I'm on windows so couldn't use it anyway but sell it to us, what's the advantages?

From what I gather they want to make something similar to Voice Control on steroids, because the idea is that we cannot use computers so need an AI agent to do it for us. They want us to talk to the computer and have it decide on what actions to perform in order to accomplish our goals. If properly implemented this could be situationally useful, however the agentic crap feels like a gimmick to attract investment by surfing the AI hype which is what irks me, as something like this could be much better if it just augmented the functionality provided by VoiceOver by making it possible to navigate totally inaccessible content such as video-games, images, and video, and then there are the privacy concerns because it is very unlikely that they will run a multi-modal large language model locally.