I love how one of the selling points of my phone (which was all the smart “AI” crap, one of which being the ability to transcribe a recording) completely falls apart the moment it’s not an American talking directly into the microphone… I don’t think it’s actually understood anything I’ve tried to use it for, even in person and very clear discussions. But my favourites are trying to record train announcements:
Lol that’s my new favourite video… All mine could understand was “11” and what the lift was saying
Here’s how that turned out:
Speaker 1 00:02
What’s
Speaker 2 00:04
The buttons? Hello, then, sold voice recognition technology in us the helmets. Voice
Speaker 3 00:10
Recognition technology. And left. And swollen. Ever tried voice recognition, technology. You know, they’re going to Scottish accents.
Speaker 2 00:21
11.
Speaker 3 00:23
Could you please repeat that?
Speaker 2 00:26
11. 11. 11.
Speaker 3 00:33
Could you please repeat that?
Speaker 2 00:36
11. This idea was us. 11.
Speaker 3 00:50
11 minutes, i’m sorry. Could you please repeat that sign language Johnson. 11. Please speak slowly and clearly smart house. 11. I’m sorry. Could you please repeat that? 11. If you don’t understand a line go away, back, came here in country. Oh, stop talking nerves that we’re back here in country. Can you be racist? Please speak slowly and clearly. 11. 11. 11 11.
Speaker 1 01:48
11. 11. 11.
Speaker 2 01:59
This is a voice activated elevator please state which floor, you would like to go to in a clear and calm manner.
Speaker 3 02:08
Was that common fit? Why is it telling people to be, calm? Because i knew they say, honesty Scottish people would be gone out for months. You have not selected a floor.
Speaker 2 02:20
If you would like to get out of the elevator without a selecting a floor, simply say open the doors, please Please, please. Suck my money. Open the doors. Please, please authentic, please remain calm. Just wait for it to speak. You have not selected a floor. Use again.
Speaker 1 03:07
People out of this guy just gave you a voice and i’m gonna go to latent child free. Going up.
Train announcements are pretty garbled on good days though? But yeah voice recognition is still pretty bad - struggle with it in my car. Have to try and americanise what I’m saying.
I always used to think “automatic transcription is hard, mistakes are inevitable” until I realised even the most noisy mumbly country-accented American speech gets transcribed basically perfectly on YouTube and other places.
I love how one of the selling points of my phone (which was all the smart “AI” crap, one of which being the ability to transcribe a recording) completely falls apart the moment it’s not an American talking directly into the microphone… I don’t think it’s actually understood anything I’ve tried to use it for, even in person and very clear discussions. But my favourites are trying to record train announcements:
“Now arriving at Berwick [Berrick]” > “now I’m riding ash ferric”
"Next [unintelligible] is Narre Warren station " > “best mattresses marijuana station”
“Berwick [Berrick]” > “Spheric”
And bonus from Sydney: “Saint Peters” became “simp penis”
I’ve always had a thing for Ash Ferric.
i love that there is a train of the month
Scottish men in voice recognition elevator https://m.youtube.com/watch?v=NMS2VnDveP8
Lol that’s my new favourite video… All mine could understand was “11” and what the lift was saying
Here’s how that turned out:
Speaker 1 00:02 What’s
Speaker 2 00:04 The buttons? Hello, then, sold voice recognition technology in us the helmets. Voice
Speaker 3 00:10 Recognition technology. And left. And swollen. Ever tried voice recognition, technology. You know, they’re going to Scottish accents.
Speaker 2 00:21 11.
Speaker 3 00:23 Could you please repeat that?
Speaker 2 00:26 11. 11. 11.
Speaker 3 00:33 Could you please repeat that?
Speaker 2 00:36 11. This idea was us. 11.
Speaker 3 00:50 11 minutes, i’m sorry. Could you please repeat that sign language Johnson. 11. Please speak slowly and clearly smart house. 11. I’m sorry. Could you please repeat that? 11. If you don’t understand a line go away, back, came here in country. Oh, stop talking nerves that we’re back here in country. Can you be racist? Please speak slowly and clearly. 11. 11. 11 11.
Speaker 1 01:48 11. 11. 11.
Speaker 2 01:59 This is a voice activated elevator please state which floor, you would like to go to in a clear and calm manner.
Speaker 3 02:08 Was that common fit? Why is it telling people to be, calm? Because i knew they say, honesty Scottish people would be gone out for months. You have not selected a floor.
Speaker 2 02:20 If you would like to get out of the elevator without a selecting a floor, simply say open the doors, please Please, please. Suck my money. Open the doors. Please, please authentic, please remain calm. Just wait for it to speak. You have not selected a floor. Use again.
Speaker 1 03:07 People out of this guy just gave you a voice and i’m gonna go to latent child free. Going up.
Train announcements are pretty garbled on good days though? But yeah voice recognition is still pretty bad - struggle with it in my car. Have to try and americanise what I’m saying.
I always used to think “automatic transcription is hard, mistakes are inevitable” until I realised even the most noisy mumbly country-accented American speech gets transcribed basically perfectly on YouTube and other places.