Topic: Speech to text as intercom display (Read 6302 times)

Scott Helmke · « **on:** October 28, 2018, 11:15:50 PM »

Had a totally goofy but would-love-to-have idea today.

What if you took a phone, tablet, etc., had it running a speech to text app, and hooked it up to listen to the show intercom?

No need for a flasher, no worries if you were away from your station, just look and see if there's any text there. You don't need to wear a headset to catch cues, either.

Are we there yet? Anybody already doing this?

Mike Caldwell · « **Reply #1 on:** October 29, 2018, 12:05:06 AM »

While my work does not require a lot of intercom usage that's not such a "goofy" idea. If your in an area where you just can't talk or make any sound you even take it one step further and have a speech synth response from a typed response, type a response and hit send.

That said there's probably already some instant messaging service that could be adapted to do just that.

Scott Helmke · « **Reply #2 on:** October 29, 2018, 08:45:36 AM »

I already know guys who already have laptops set up (speaker control, etc) and use that to message back and forth. I'm thinking more about corporate type shows where I need to mix music but also need to catch the occasional cue. So I'm thinking about something that could have about the same footprint and connection as a flasher.

Jason Glass · « **Reply #3 on:** October 29, 2018, 06:22:33 PM »

Quote from: Scott Helmke on October 28, 2018, 11:15:50 PM

Had a totally goofy but would-love-to-have idea today.

What if you took a phone, tablet, etc., had it running a speech to text app, and hooked it up to listen to the show intercom?

No need for a flasher, no worries if you were away from your station, just look and see if there's any text there. You don't need to wear a headset to catch cues, either.

Are we there yet? Anybody already doing this?

Wow! I would LOVE this in high noise environments. This is a Pete Erskine subject if there ever was one!

Sent from my mobile phone using Tapatalk

Scott Helmke · « **Reply #4 on:** October 29, 2018, 06:30:04 PM »

This morning I spent maybe a minute playing around with an old Android tablet and the built-in speech to text. It's pretty limited in what it does, with long delays while it tries to decide what you said and maybe waiting for more before deciding. And of course it's expecting just a sentence or two and then turns itself back off, where we'd need an app that just stayed open and accumulated/scrolled text for hours at a time.

Lou Kohley · « **Reply #5 on:** October 29, 2018, 06:59:58 PM »

While we are on the subject is there a simple messaging setup that can be used without an internet connection. Basically an instant messenger on a closed network.

LOU

Scott Holtzman · « **Reply #6 on:** October 29, 2018, 07:24:56 PM »

Quote from: Lou Kohley on October 29, 2018, 06:59:58 PM

While we are on the subject is there a simple messaging setup that can be used without an internet connection. Basically an instant messenger on a closed network.

LOU

Yes, I would run ircd or xmppd on a Raspberry Pi. Almost any chat client can connect to an xmpp server. IRCD you will have to use and IRC client.

Edited for clarity. Those are both servers. You would then connect your chat client on whatever device (computer, phone, tablet) to the local server (The Raspberry Pi) instead of a server on the Internet.

Lou Kohley · « **Reply #7 on:** October 29, 2018, 07:30:04 PM »

Quote from: Scott Holtzman on October 29, 2018, 07:24:56 PM

Yes, I would run ircd or xmppd on a Raspberry Pi. Almost any chat client can connect to an xmpp server. IRCD you will have to use and IRC client.

Edited for clarity. Those are both servers. You would then connect your chat client on whatever device (computer, phone, tablet) to the local server (The Raspberry Pi) instead of a server on the Internet.

Thanks for the tip. Ill get reading up on it.

LOU

Andrew Broughton · « **Reply #8 on:** October 30, 2018, 12:45:23 PM »

Interesting idea, but I would guess that speech recognition would have to get a lot better, if my (sometimes hilarious) visual voicemail messages are any indication of the state of speech recognition. Siri and Alexa barely work and they're just looking for simple phrases.
Now add into that all the background noise of a typical show, I don't think there'd be much hope of anything intelligible at the output of the recognition system... And lots of false text where there isn't anyone speaking because someone left their mic open.

Dave Garoutte · « **Reply #9 on:** October 30, 2018, 12:58:41 PM »

The promoter at one of my gigs sends texts up to the booth when he wants some tweaks.
I think having displayed text would be great, as it would stay there until you respond, so if you're deep into something you can still see the message when you're done.
That said, with auto 'correct' I can see some serious potential issues; 'I need more buffalo in the drone monkey'.

Erik Jerde · « **Reply #10 on:** October 30, 2018, 03:02:05 PM »

You could probably feed I put from an audio interface into a product like dragon naturally speaking. If memory serves that’s a much more real-time transcription software. The problem may come in feeding it audio from multiple voices that it hasn’t been trained on in a less than ideal environment.

Scott Holtzman · « **Reply #11 on:** October 30, 2018, 03:08:15 PM »

Quote from: Erik Jerde on October 30, 2018, 03:02:05 PM

You could probably feed I put from an audio interface into a product like dragon naturally speaking. If memory serves that’s a much more real-time transcription software. The problem may come in feeding it audio from multiple voices that it hasn’t been trained on in a less than ideal environment.

Dragon is more an app. would you have Internet access? Amazon Polly works pretty well. https://aws.amazon.com/polly/

I have not played with Google https://cloud.google.com/speech-to-text/

The cloud based services are much more powerful than a single PC could be.

Scott Helmke · « **Reply #12 on:** October 30, 2018, 03:51:47 PM »

I'd imagine if I was waiting for "House to half" I could figure out "How's to Have?" on the screen.

Randy Pence · « **Reply #13 on:** October 30, 2018, 07:07:15 PM »

google keep is a free app which offers a speech to text function and the corresponding transcription can be sent as a note to other users. Ive messed around with both it and the web browser version and the web one worked a lot better than mobile on that particular day.

sending the keep notes is probably not as user friendly to quickly communicate to foh, but might work

Scott Holtzman · « **Reply #14 on:** October 30, 2018, 10:44:32 PM »

Quote from: Randy Pence on October 30, 2018, 07:07:15 PM

google keep is a free app which offers a speech to text function and the corresponding transcription can be sent as a note to other users. Ive messed around with both it and the web browser version and the web one worked a lot better than mobile on that particular day.

sending the keep notes is probably not as user friendly to quickly communicate to foh, but might work

With the API's on the Amazon and the Google service you could hack a little bit of code (Python is as easy as BASIC was for our generation). Grab the audio from the input, send it to the cloud and shove the results into the XMPP server for display in the chat app. https://realpython.com/python-speech-recognition/

Tim Hite · « **Reply #15 on:** October 30, 2018, 11:24:48 PM »

Apple iChat used to connect across the LAN without going overseas. There are loads of other LAN chat applications, here's a simple one. . .

http://lanmsngr.sourceforge.net

Quote from: Lou Kohley on October 29, 2018, 06:59:58 PM

While we are on the subject is there a simple messaging setup that can be used without an internet connection. Basically an instant messenger on a closed network.

LOU

Lou Kohley · « **Reply #16 on:** October 31, 2018, 01:49:38 AM »

Quote from: Tim Hite on October 30, 2018, 11:24:48 PM

Apple iChat used to connect across the LAN without going overseas. There are loads of other LAN chat applications, here's a simple one. . .

http://lanmsngr.sourceforge.net

Thanks Tim.
Simple. Just what I needed.

I did read up on xmpp servers and it seems like a fairly easy process. This will be a stepping stone on the way to building the pi and making a server...one day.

LOU

Andrew Broughton · « **Reply #17 on:** October 31, 2018, 01:14:54 PM »

Quote from: Tim Hite on October 30, 2018, 11:24:48 PM

Apple iChat used to connect across the LAN without going overseas. There are loads of other LAN chat applications, here's a simple one. . .

http://lanmsngr.sourceforge.net

Is there one out there that supports all platforms (MacOS/Windows/Linux/iOS/Android)

Riley Casey · « **Reply #18 on:** October 31, 2018, 02:07:49 PM »

Lets think about this for a minute. Speech to text is often ridiculously bad even when used in good acoustic environments with little or no competing noise let alone competing speech. More often than not the person transmitting information is in the same acoustic environment as the show meaning crowd noise, loud music and probably worst of all for S to T applications competing speech from the program. While I can certainly understand the usefulness of text messaging in its various forms between show tech crew I can't really see sending go messages to talent or video crews via text or calling spot pick ups.

https://www.youtube.com/watch?v=JepKVUym9Fg

Quote from: Jason Glass on October 29, 2018, 06:22:33 PM

Wow! I would LOVE this in high noise environments. ...

Tim Hite · « **Reply #19 on:** October 31, 2018, 03:10:24 PM »

http://lanmsngr.sourceforge.net/downloads.php

Looks like LAN Messenger handles the big 3. I would think you could easily recompile the source code to handle android. iOS is another matter. Maybe ask the developer for help?

Quote from: Andrew Broughton on October 31, 2018, 01:14:54 PM

Is there one out there that supports all platforms (MacOS/Windows/Linux/iOS/Android)

Pete Erskine · « **Reply #20 on:** December 09, 2018, 02:41:45 AM »

Quote from: Scott Helmke on October 28, 2018, 11:15:50 PM

Had a totally goofy but would-love-to-have idea today.

What if you took a phone, tablet, etc., had it running a speech to text app, and hooked it up to listen to the show intercom?

No need for a flasher, no worries if you were away from your station, just look and see if there's any text there. You don't need to wear a headset to catch cues, either.

Are we there yet? Anybody already doing this?

YES... check out http://avlifesavers.com/comtext.htm

ProSoundWeb Community · Â« **Reply #20 on:** December 09, 2018, 02:41:45 AM Â»

ProSoundWeb Community

Author Topic: Speech to text as intercom display (Read 6302 times)

ProSoundWeb Community