• Our booking engine at tickets.railforums.co.uk (powered by TrainSplit) helps support the running of the forum with every ticket purchase! Find out more and ask any questions/give us feedback in this thread!

The abhorrent Text to Speech system - making the traveling experience worse?

Status
Not open for further replies.

RailWonderer

Established Member
Joined
25 Jul 2018
Messages
1,588
Location
All around the network
I've previously only heard TTS in Kent and I've heard Scotrail are getting it as well. It has since arrived in East Anglia at Chelmsford and Colchester (but not the smaller stations yet). ToCs are switching to this robotic monotone voice for 'flexibility' but why does a voice have to say anything other than what the live feed has to say? Station staff can add a message if need be.
Don't rail companies consider this part of the traveling experience or do they believe convenience of saying anything they want it to say is all that matters and passengers won't care?
The voice cannot even speak properly.
Calling at Ingate STONE, ShenfieldStratford and LONdonLiverpoolSt.
It would improve the traveling experience to have someone who sounds like Phil with the four chimes before the baritone voice.

The new PIS screens that accompany them are rubbish as well, two minutes after a train has left it still displays that service.
 
Sponsor Post - registered members do not see these adverts; click here to register, or click here to log in
R

RailUK Forums

Bletchleyite

Veteran Member
Joined
20 Oct 2014
Messages
97,534
Location
"Marston Vale mafia"
I think they should work with Aviavox about doing a TTS system that uses actual voice snippets (syllables, or what they call "fragments") to produce a clearer, more human voice, as used in most airports.
 

ComUtoR

Established Member
Joined
13 Dec 2013
Messages
9,398
Location
UK
I've previously only heard TTS in Kent and I've heard Scotrail are getting it as well. It has since arrived in East Anglia at Chelmsford and Colchester (but not the smaller stations yet). ToCs are switching to this robotic monotone voice for 'flexibility' but why does a voice have to say anything other than what the live feed has to say? Station staff can add a message if need be.

You need a voice because not everyone can read.

Don't rail companies consider this part of the traveling experience or do they believe convenience of saying anything they want it to say is all that matters and passengers won't care?

I would say its because passengers care. The old pre-canned 'excuses' are no longer suitable, neither are they desired. Passengers want more 'human like' messages and they want their information up to date, open and honest. To achieve that you need flexibility in the system. You also need something that can be controlled centrally for consistency across multiple locations and you also need something.. Dare I say.. Cheap.
 

Dai Corner

Established Member
Joined
20 Jul 2015
Messages
6,316
You need a voice because not everyone can read.



I would say its because passengers care. The old pre-canned 'excuses' are no longer suitable, neither are they desired. Passengers want more 'human like' messages and they want their information up to date, open and honest. To achieve that you need flexibility in the system. You also need something that can be controlled centrally for consistency across multiple locations and you also need something.. Dare I say.. Cheap.

I'm not familiar with the system, but would it be better if the operator spoke the message and it was recorded to be played out as necessary? I'm assuming reasonably good, clear voices.
 

RailWonderer

Established Member
Joined
25 Jul 2018
Messages
1,588
Location
All around the network
I would say its because passengers care. The old pre-canned 'excuses' are no longer suitable, neither are they desired. Passengers want more 'human like' messages and they want their information up to date, open and honest. To achieve that you need flexibility in the system.
How the message is said is just as important as what is being said. A slightly more specific train announcement (e.g. The xx:xx to London Liverpool St has been cancelled. This is due to a trampoline on the line. Rather than: This is due to an obstruction on the line). The former is slightly more specific and the latter is said in a human voice. Why do passengers need a more precise reason for a delay or cancellation, what difference will it make? The humanity in it is more important.
 

ComUtoR

Established Member
Joined
13 Dec 2013
Messages
9,398
Location
UK
Why do passengers need a more precise reason for a delay or cancellation, what difference will it make?

A million RUkf Members just screamed out in pain....

Seriously, on the ground it doesn't make much of a difference and I find it sometimes adds to the problems but there is a part of the traveling public that craves every single drop of information. Keyboard warriors take to Twitter and hit the forums. The angry mob scream and shout at staff and the rest just huff and puff to themselves. Just the other day, on this very forum, a TOC was accused of lying about a delay.

Could the system be better ? Of course and no doubt it will be improved over time.
 
Last edited:

_toommm_

Established Member
Joined
8 Jul 2017
Messages
5,843
Location
Yorkshire
Didn't Arriva Trains Wales use something very similar? I remember sitting at Helsby and hearing a rather funny announcement about a stopping service to Holyhead. It also split and went to Llandudno which was quite painful to listen to.
 

Bletchleyite

Veteran Member
Joined
20 Oct 2014
Messages
97,534
Location
"Marston Vale mafia"
Didn't Arriva Trains Wales use something very similar? I remember sitting at Helsby and hearing a rather funny announcement about a stopping service to Holyhead. It also split and went to Llandudno which was quite painful to listen to.

Yes, at quieter stations their displays have a speaker which read out in TTS what's on display.

Given how good some TTS is e.g. Alexa (or Aviavox), I think there's no excuse for doing it on the cheap.
 

cactustwirly

Established Member
Joined
10 Apr 2013
Messages
7,447
Location
UK
Didn't Arriva Trains Wales use something very similar? I remember sitting at Helsby and hearing a rather funny announcement about a stopping service to Holyhead. It also split and went to Llandudno which was quite painful to listen to.

EMR use it at the vast majority of their stations
 

RailAleFan

Member
Joined
2 Jul 2014
Messages
314
Location
Midlands
This was a Friday afternoon project from some time ago and a bit rough around the edges as the Web Speech API isn't consistently or that well supported yet but gives some idea of current OS built-in text to speech capabilities with real world and real time station announcements;

https://live-departures.info/rail/dsa/

Most operating systems offer a selection of voice engines - on Mac OS try "Fiona (en-scotland)" for locations in Scotland which of what I have tried so far is the most natural sounding (if a bit fast).

No phonetic translations are in place so the text on screen is exactly as passed to the API (so "Reading" will be announced as in "Reading a book")
 
Last edited:

bishdunster

Member
Joined
9 Oct 2012
Messages
278
Location
Dunster
Try typing in Woking, it is bloody abysmal; i speak as a former B.R. station announcer (Bournemouth 1970s)o_O !
 

causton

Established Member
Joined
4 Aug 2010
Messages
5,504
Location
Somewhere between WY372 and MV7
Yes, at quieter stations their displays have a speaker which read out in TTS what's on display.

Given how good some TTS is e.g. Alexa (or Aviavox), I think there's no excuse for doing it on the cheap.

I heard somewhere that the Greater Anglia screens (and the WMT ones, as they will, one day, use the same system) use Amazon's text-to-speech, or at least will/should do...
 

High Dyke

Established Member
Joined
1 Jan 2013
Messages
4,267
Location
Yellabelly Country
EMR use it at the vast majority of their stations
Aye...good old Doris! Places like Thurgarton are said as if it was such a dirty word. The ability to frequently waffle inanely about security, especially on a country station with an hourly train service; oh and advising passengers to take care due to adverse weather, even if the weather at your location anytime of day, or night, is perfectly dry.
 

daveshah

Member
Joined
1 Sep 2018
Messages
115
No phonetic translations are in place so the text on screen is exactly as passed to the API (so "Reading" will be announced as in "Reading a book")

Impressively with the default settings on Android Chrome, Reading is pronounced correctly, I can only presume it picks up on the capital R. Other than saying "colon" for the colons in times it actually does pretty well!
 

Peter Kelford

Member
Joined
29 Nov 2017
Messages
903
I've previously only heard TTS in Kent and I've heard Scotrail are getting it as well. It has since arrived in East Anglia at Chelmsford and Colchester (but not the smaller stations yet). ToCs are switching to this robotic monotone voice for 'flexibility' but why does a voice have to say anything other than what the live feed has to say? Station staff can add a message if need be.
Don't rail companies consider this part of the traveling experience or do they believe convenience of saying anything they want it to say is all that matters and passengers won't care?
The voice cannot even speak properly.
Calling at Ingate STONE, ShenfieldStratford and LONdonLiverpoolSt.
It would improve the traveling experience to have someone who sounds like Phil with the four chimes before the baritone voice.

The new PIS screens that accompany them are rubbish as well, two minutes after a train has left it still displays that service.
In France, the new-ish 'eMone' system works by recording individual words and merging them into phrases and sentences.

Impressively with the default settings on Android Chrome, Reading is pronounced correctly, I can only presume it picks up on the capital R. Other than saying "colon" for the colons in times it actually does pretty well!
I've played around with TTS before. If you want it to say TER in the sense of the French regional train and not in the pronunciation of 'terre', you would write T-E-R. Likewise, for Reading, you would write Redding etc.
 

Bletchleyite

Veteran Member
Joined
20 Oct 2014
Messages
97,534
Location
"Marston Vale mafia"
I've played around with TTS before. If you want it to say TER in the sense of the French regional train and not in the pronunciation of 'terre', you would write T-E-R. Likewise, for Reading, you would write Redding etc.

Yes, if I was implementing a system I'd have a table of "word", "context" (e.g. station, noun, verb etc) and "phonetic translation" in the database. You could then easily fix ones that sound wrong.

FWIW, this principle can be used to get Alexa to swear :D
 

James H

Member
Joined
25 Jun 2014
Messages
1,083
I do think it’s a real problem for the railways in terms of passenger experience (especially when the TTS system is poor).

The message is gives is that “we don’t care enough to have a real person communicate important information about your journey”.

The worst systems sound dystopian.

The system used by LU at Kings Cross sounds dreadful.

Likewise the Southeastern announcements heard at London Bridge and elsewhere.
 

py_megapixel

Established Member
Joined
5 Nov 2018
Messages
6,642
Location
Northern England
I've previously only heard TTS in Kent and I've heard Scotrail are getting it as well. It has since arrived in East Anglia at Chelmsford and Colchester (but not the smaller stations yet). ToCs are switching to this robotic monotone voice for 'flexibility' but why does a voice have to say anything other than what the live feed has to say? Station staff can add a message if need be.
Don't rail companies consider this part of the traveling experience or do they believe convenience of saying anything they want it to say is all that matters and passengers won't care?
The voice cannot even speak properly.
Calling at Ingate STONE, ShenfieldStratford and LONdonLiverpoolSt.

Those dreadful screens are from a company called Blackboxco which is full of screenshots of the displays working perfectly. But from what I've heard, the reality seems to be anything but.

I'm unsure what the TTS system in Kent is, but Abellio's text-to-speech engine is provided by Amazon - it seems to be IVONA - a legacy system so-called because it was the name of a company taken over by Amazon. It could be Polly, which is Amazon's TTS system which succeeded IVONA, but the demo voices on Polly sound much more natural, and completely different to the voices on Greater Anglia platforms, so it would have to be an incredibly bad implementation of Polly. (Sorry if that doesn't sound clear; it's a summary of a long paragraph I wrote over on the Automated Announcements forum).


Yes, if I was implementing a system I'd have a table of "word", "context" (e.g. station, noun, verb etc) and "phonetic translation" in the database. You could then easily fix ones that sound wrong.

FWIW, this principle can be used to get Alexa to swear :D

Indeed; playing around with TTS in the past I've found that it's much easier to have two separate words stored: what will be displayed on the screen, and what will be read out. The problem is that even with that modification, it will still sound unnatural, and this makes listening to it for any time incredibly grating.


I do think it’s a real problem for the railways in terms of passenger experience (especially when the TTS system is poor).

The message is gives is that “we don’t care enough to have a real person communicate important information about your journey”.

The worst systems sound dystopian.

The system used by LU at Kings Cross sounds dreadful.
I assume you mean the "There is severe disruption on the Hammersmith and City line.... there is a good service on all other lines" thing? Because the approaching train announcements are pre-recorded and sound very natural in my opinion.
 

Bletchleyite

Veteran Member
Joined
20 Oct 2014
Messages
97,534
Location
"Marston Vale mafia"
Indeed; playing around with TTS in the past I've found that it's much easier to have two separate words stored: what will be displayed on the screen, and what will be read out. The problem is that even with that modification, it will still sound unnatural, and this makes listening to it for any time incredibly grating.

I'm not overly convinced the level of wittering on we have is necessary anyway. It's done primarily for accessibility reasons, but surely we can't be far off the point where blind or partially sighted people might prefer to choose to carry a mobile phone which could be designed to automatically provide such information to them on demand rather than bombarding everyone with it.
 

RailAleFan

Member
Joined
2 Jul 2014
Messages
314
Location
Midlands
Impressively with the default settings on Android Chrome, Reading is pronounced correctly, I can only presume it picks up on the capital R. Other than saying "colon" for the colons in times it actually does pretty well!
Wow - i'd not tried it in Android/Chrome for a long time - that's come on a long way. System default is the same voice as Google Assistant on my Pixel 3.

I'm surprised however at it not identifying HH:MM as a time which I'm sure used to be the case. I was planning on standardising this anyway as some voice engines convert times into an alternative format (possibly depending on locale settings) for example 15:15 being spoken as "three fifteen PM". Just applied a quick mod to normalise that.
 

Bletchleyite

Veteran Member
Joined
20 Oct 2014
Messages
97,534
Location
"Marston Vale mafia"
What's wrong with the old screens, the dot matrices for the general public and the black screens with blue borders for anything slightly more staff orientated?

Not an awful lot, and I think Abellio are wasting an immense amount of money by replacing technology that works just fine with technology that is in many ways inferior.

Though they do need to replace the one on Bletchley P4/5 which has had a dud line for about 3 years now.
 

RailWonderer

Established Member
Joined
25 Jul 2018
Messages
1,588
Location
All around the network
Not an awful lot, and I think Abellio are wasting an immense amount of money by replacing technology that works just fine with technology that is in many ways inferior.
This comes back to new or rebranding ToCs who want to give the illusion of change and investment in railways by rejigging the look of the stations with those black-box.co boards that are actually cheaper (according to another thread here) when the old dot matrices were absolutely fine. Style over substance and form over function is the only way to put it.
The message is gives is that “we don’t care enough to have a real person communicate important information about your journey”.
Absolutely. Everything in the railways these days seems is PRM - convenience obsessed, from seating to the announcements to those ugly foot bridges they are installing at hundreds of stations. Truth is the TTS often sounds dreary, distopian and not a place that makes the railway pleasant for people. There was nothing more welcoming than hearing the four chimes and Phil’s announcement following.

There was an interview with TfL’s announcer (the woman who does the buses and LO services) who said she recorded her lines with a smile because she wanted to sound slightly friendly but not too patronising or dull at the same time, for commuters who may be miserable on their daily journey. Phil was the perfect voice for that reason and the railway experience is made up of getting small details right that are not abysmally expensive to reproduce.
 
Status
Not open for further replies.

Top