DanielTheEMid
Member
- Joined
- 13 Mar 2019
- Messages
- 103
Yeah, Real Vanessa vs Readspeaker's TTS Vanessa, which I think a combination of nonsense sentences and stations to grab every phoneme in the english language was required to be able to voicebank.
-
Our new ticketing site is now live! Using either this or the original site (both 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!
New ScotRail Train Announcements
- Thread starter dr140
- Start date
- Status
Sponsor Post - registered members do not see these adverts; click here to register, or click here to log in
R
RailUK Forums
reecestrains
Member
I feel like I've heard the voice before somewhere...
TT-ONR-NRN
Veteran Member
That's definitely "Scottish Female"
So we finally have a name for her, Vanessa!
Unfortunately, although I actually love her platform announcements, and am one of the few people who controversially prefer them to Allison McKay's, I do think these on-board ones sound awful.
Indeed, they really do sound like the unknown staff member that makes the platform announcements throughout the central and southern areas. Sadly, TTS has distorted her voice quite a bit so it doesn’t sound great. But at least we can refer to ‘Scottish Female’ as ‘Vannessa’ now, (from Paisley).
Buzby
Established Member
I really cannot detect any real improvement - just another Scottish female doing TTS, it just seems another waste of money/time/effort (take your pick)!
tfw756rider
On Moderation
The video doesn't work for people who don't have TikTok - please can you (or @vuzzeho who knows how to do it) embed it to a post in this thread?
I have downloaded it here and attached a copy.
View attachment ssstik.io_1765485184924.mov
This should enable you to directly play the recording.
tfw756rider
On Moderation
Many thanks! Yeah, they could do with a little more work I say
After listening to this i agree it very likely is Scottish Female from the Zenitel system voicing it. The voice and the way some words are said sounds very similar. So i think that Vanessa Sloan is indeed her.
Scottish Female actually sounds really good on the station Zenitel system. I have always considered her to be one of the better staff members voice. She has a nice sounding voice.
But this new TTS version is awful. Putting her through TTS software has messed up her voice. The software has made some words like "Cardonald" and "Fairlie" just sound really weird. Her voice goes all weird and makes her sound like she is either drunk or unwell. It sounds so bad.
If they really want to save money and replace Fletcher Mathers then they could have got Vanessa Sloan to record proper announcements like she does for her station Zenitel station announcements. Creating a TTS voice just sounds bad quality. It has messed up her voice but in real life she actually has a nice sounding voice.
Scottish Female actually sounds really good on the station Zenitel system. I have always considered her to be one of the better staff members voice. She has a nice sounding voice.
But this new TTS version is awful. Putting her through TTS software has messed up her voice. The software has made some words like "Cardonald" and "Fairlie" just sound really weird. Her voice goes all weird and makes her sound like she is either drunk or unwell. It sounds so bad.
If they really want to save money and replace Fletcher Mathers then they could have got Vanessa Sloan to record proper announcements like she does for her station Zenitel station announcements. Creating a TTS voice just sounds bad quality. It has messed up her voice but in real life she actually has a nice sounding voice.
scottishtrains
Member
Thanks for doing that, I was going to but couldn’t figure out how
applepie2100
Member
- Joined
- 16 Aug 2011
- Messages
- 170
It still sounds absolutely terrible. It seems like a cheap and nasty solution instead of actually spending the money and doing it properly. I'm sure there are better quality text to speech solutions available but they're not willing to spend the money on them. They resolved the issue of the original controversial recordings but not the issue of the quality.
If they want a Scottish voice they should stick with Fletcher Mathers and do it through recordings as previously done or go with one of the voices used down south.
If they want a Scottish voice they should stick with Fletcher Mathers and do it through recordings as previously done or go with one of the voices used down south.
Having now heard it in the flesh, it's certainly an improvement over the fully AI voice, which I found totally grating. The new voice obviously sounds disjointed due to the way the sentences are formed, but she is perfectly clear to me. I would agree that the original Fletcher Matthew's voice was still the best though.
scottishtrains
Member
Same here, it doesn’t seem that bad
318266
Member
If you don't know what you are doing when setting up a system largely trained on general American announcement voices (which I wager is true in the case of Readspeaker), I would not be confident for it to deal with a Scottish voice, given the surprising complexity in a Scottish voicebank compared to a general American (say New England) accent. We now have a Sonya AI built upon the voice of someone who is otherwise quite well articulated, as seen with the station announcements.
I question why the AI is necessary for every single announcement, instead of just the ones that apparently need to be generated on-the-fly (on that note, does the present 380 system, as quirky and oddly-faulty as can be, even work with such needs? And why couldn't conductors perform such an announcement manually if necessary?).
I question why the AI is necessary for every single announcement, instead of just the ones that apparently need to be generated on-the-fly (on that note, does the present 380 system, as quirky and oddly-faulty as can be, even work with such needs? And why couldn't conductors perform such an announcement manually if necessary?).
D6130
Established Member
The majority of trains in the Strathclyde area don't have conductors, IIRC.
Buzby
Established Member
I suppose it depends on how you define the task, but on all my trips they walk through, checking (and selling) tickets as appropriate whilst operating the doors and giving the proceed bell to the drive. The trains usually run with a staff of two.
hexagon789
Veteran Member
On guard-operated trains yes, thats how they work.
On the DOO (driver-only operated) services, the second staff member does ticket checks/sales and nothing else.
DOO applies on Inverclyde, Ayrshire, Cathcart/Newton/Neilston, Argyle Line, North Clyde for instance.
EK/Barrhead remain with guards, even with the 380s.
I believe the station ones are actually recordings as opposed to being a Text to Speech.
They are, if you dig for them; you could probably find them through an FOI or otherwise, but I'm not sure.
I get why they've put them into TTS, for the added flexibility, but I wonder if they could've just asked the original voice artist?
== Doublepost prevention - post automatically merged: ==
Also, there are just some weird gaps in between words that do not sound right. It just sounds 'robotified'
Last edited:
Yes all of the station files voiced by her were released by Network Rail under the FOI act:
The station recordings are proper traditional (not TTS) recordings but the voice is the same lady (Vannessa Sloan) that was used to create the new TTS onboard voice.
I reckon they are going for TTS because it is easier and quicker to update. If they need a new recording then they can just generate it instantly. But if this was proper recordings (like the stations ones are) then they would have to wait for her to go back to the studio and record it. But of course the downside is that TTS sounds far less natural and words do not flow as well.
- Status