THIS DESTROYS Apple and Google's STT

This is a two parter: a demo of why Siri and Google Assistant can't even come close to OpenAI's Whisper, and then a walkthrough of running Whisper on your machine.
👀 Gear I use: www.amazon.com/shop/alexziskind
▶️ Setting up new M2 Mac Mini for software development - • Setting up new M2 Mac ...
▶️ python environment setup - • python environment set...
▶️ AWS UNLEASHED their Copilot killer - • AWS UNLEASHED their Co...
▶️ How to get Copilot FREE - • Get GitHub Copilot for...
▶️ github gave me a BEAST for coding - • github gave me a BEAST...
▶️ Apple M1 vs Intel i9 JavaScript Nx Workspace Install and Build Speed Comparisons - • Apple M1 vs Intel i9 J...
#programming #ai #softwaredevelopment
💻NativeScript training courses - nativescripting.com
(Take 15% off any premium NativeScript course by using the coupon code YT2020)
👕👚iScriptNative Gear - nuvio.us/isn
- - - - - - - - -
❤️ SUBSCRIBE TO MY KZbin CHANNEL 📺
Click here to subscribe: / alexanderziskind
- - - - - - - - -
🏫 FREE COURSES
NativeScript Core Getting Started Guide (Free Course) - nativescripting.com/course/na...
NativeScript with Angular Getting Started Guide (Free Course) - nativescripting.com/course/na...
Upgrading Cordova Applications to NativeScript (Free Course) - nativescripting.com/course/up...
- - - - - - - - -
📱LET'S CONNECT ON SOCIAL MEDIA
ALEX ON TWITTER: / digitalix
NATIVESCRIPTING ON TWITTER: / nativescripting
00:00 Demo
03:00 Whisper install walkthrough

[@AZisk]

This dude's hair is getting insane . But aside from that, do you use Siri or Google Assistant or Whisper?

[@JeanBrito666]

At least you have hair yet... hahaha

[@antenedilbert7191]

Resolve!!! Da Vinci!!!!!!!! Blazingggggggggggggggggg Fasttttttttttttttttttttt

[@mjlynden]

This looks like my ticket to transcribing foreign tv shows and movies that don’t have English subtitles yet. I see there is options to translate the transcribe and you even had an srt subtitle file spat out which would have timecodes already! Amazing!

[@harrisontu264]

Oh, that’s interesting. I’d be really curious if error rate would go up though, just due to other sounds in the movie and things like that. I don’t know how good whisper translate is, but if it sucks, you could always write a deepl srt translator. Actually, not a bad idea in any case.

[@OpulentWonders]

I've been using the paid version provided by Openai through an API. I found it's slightly more accurate than the version I can run locally due the ability to provide a brief summary of the video. Not to mention faster. I would love to see the .cpp version on this channel though!

[@vaisakhkm783]

It is really useful to transcribe entire collage class and company meetings, make a script to download it with ytdl, excract audio with ffmpg, and pipe it to this :) later give it to some ai summarization bot to write note... time stamp included, so if there is any doubt while reading summary, i can go to that time

[@yzengame]

Very helpful. Thanks as always.

[@MrZiemwit]

thic chanel is so good, wow 2023 and something not stupid in this world :)

[@dudelkt]

Awesome content as always!

[@AZisk]

thx 🙏

[@FilmFactry]

When it gets easier, I will use it. I think it would be a very special use case to set all this up for a transcript.

[@mohamadkadabashi]

Yes, please do the version with c++. I had some issues to install python 2.7. How did you do that? And how did you switch between the versions ?

[@harrisontu264]

One idea I’ve been kicking around, for a while is to combine whisper plus gpt4all plus whatever the systems native speech output is, to create a local, on device, chat bot, that you could actually talk to. No cloud necessary.

[@ChrisGVE]

Potential to transcribe meeting minutes… I wonder if it recognize different voices in a given recording

[@AZisk]

whisper by default doesn’t recognize different voices, but there is a plugin that does, but i couldn’t get it to work.

[@SlosII]

could you use this to generate subtitles (I noticed it created a SRT file) - and can these subtitles be automatically translated to other languages? That'd be awesome.

I have a lot of classes recorded from the pandemic isolation that I would like do transcribe, with the goal of creating subtitles. So I think that whisper my be a good tool for educational purposes. There an AWS service, called Transcribe, that is kind of good too.

[@HadesTimer]

This would be great for closed captions on KZbin videos. I see so many people looking for accurate CC.

[@AnthonyBrown-ww5sn]

This could be used for transcribing meetings of all types.

[@simonebrazzi287]

Hi Alex 🙂 Have you ever tried Notion AI? It should be the cheapest but also seems very well done. A plus for me, it is integrated with Notion. Maybe a video...

[@iddeegee]

2 scenarios, for interviews or something similar, to have a text version.. or for song lyrics (when you couldn't find a text) But not sure if I would use it for notes on the go, as it's PC/Mac dependant, but for converting old records, is a way to go

[@harrisontu264]

I don’t know if you’re an iOS user, but that aiko app he mentioned is available for iOS just so you know. I’ve never used it, but I would assume it would be pretty good because I’ve use the models before and they are very good.

[@darkbelg]

The use i see for the whisper model locally is to replace "hi siri". So you run the tiny model locally every second or every 5 seconds and then when you actually want dictation or talk to chatGPT you use the whisper api. This keeps cost low and also resources locally also. I don't know how fast python can scale up or load another model. I don't know how fast the smallest whisper model can be changed to the larger one.

[@shahriar0247]

you might wanna try google's recorder, i installed a pixel rom on a oneplus phone, even tho i didnt have the tensor chip, it was amazing, like maybe a tiny little slow, but i didnt see any perfrmance differnt to real tensor, the speech to text i mean

[@ZodiToo-px6jj]

Hey @AZisk, I like your videos especially those about using Visual Studio on a Mac. Can you make a video about how to use Visual Studio on Windows in Parallels on a Mac to debug Android apps? I want to be able to choose an Android VM on my Mac while using Visual Studio on Windows in Parallels. I tried to figure it out but it’s hard. Thanks for your videos and keep making them!

[@vaisakhkm783]

Google's livetranscribe app can go for hours... but potential privacy concers are still there

[@a-terrible-fate532]

lets get whisper to be our voice typing on our smart phones, ultra fast and accurate !

[@hfislwpa]

Been using this to transcribe youtube videos yeah! Pretty cool

[@AZisk]

is it better than the built-in transcription that youtube provides?

[@ehza]

Hey, I just notice Bengali is not in the list. How could I train whisper for the Bengali language? Could you make a video on this? The video doesn't necessarily have to show for the Bengali language. Any language would suffice!

[@MinisterioLasJaeles]

Make a video how to use or integrate AI with a Nativescript app.

[@romcomtom9368]

I can see this being integrated into home assistant.

[@jsmsj]

nice

[@AZisk]

thx - glad you enjoyed

[@anujraymajhi812]

nice to see it performing well, but sad cause I'm from Nepal and it has highest Word error rate for Nepali, prolly because of data availability... but overall nice model

[@AZisk]

yes, i wonder have you tried it? and does it in fact perform that poorly on Nepali? And also, is Siri or Google better at Nepali?

[@jameshancock]

Keyboards that automatically correct at the end of sentences.... going to be awesome.

[@Aladinzc]

Damn. That looks awsome. Damn. I am not developer and lost in this stuff. :( Is there some similat solution for normal user? Do not need to be super powerfull. I installed some audio source hack that sends input into Audio to text engine in Mac but it does not listen continuosly so it is lame. And I no not like Google. Otherwise one can upload video to KZbin and get transcription there.

[@LuisJimenez-br1nl]

Hmmmm. You should redo the test since it's not fair to use siri on an iphone and google assistant on another phone than a pixel 7. As far as I know It works better on a pixel

[@Thekidisalright]

Aiko pronounced as “i-ko” ☺️

[@tykealy]

Just realized that PUTIN is a tech guy.

[@AZisk]

Underrated comment.

[@dannyisrael]

I don’t get it 😅

[@AZisk]

@@dannyisrael commenter thinks I look like Putin.

[@ToySeeker]

😂❤❤❤

[@danbrowne7796]

While I understand this is about Whisper you are implying it is better than Google assistant and what you completely missed and didn't mention is that speech to text works nearly flawlessly on a pixel device due to the tensor chip which unfortunately makes the video look not only biased but uninformed.

[@AZisk]

Which pixel?

[@endoflevelboss]

Everything Google makes these days is flaky as shit. So many bugs in my Nest devices. I'm done with buying their products and services.

[@zalzalahbuttsaab]

lol. I just want to press a button bud and the computer does everything else. These "just complete these several steps" videos make me laugh. I don't have the time for that. I guess this type of software is aimed at entry-level geeks rather than the average use case Joe such as myself. I'm waiting for the time when AI can do all these steps with a simple text prompt from the user.

[@AZisk]

Looks like method 1 is your friend then. The aiko app

[@zalzalahbuttsaab]

@@AZisk yes. In the modern age there should be one-button turnkey solutions for all geekery. Time for developers in their basement dungeons to get their heads out of Star Trek and into real world mass usability. Any AI would understand that. Humans need to catch up.