Some answers /Comments
Odd charecters ....
Some of them in analogue days were due to errirs/interference/ noise on the teletext Signals,,
but there were some systemic errors due to the fact that the code tables used for the Text
in EBUTech 3264
https://tech.ebu.ch/docs/tech/tech3264.pdf
which is the Subtitling data exchange format (dating from 1989)
was used by both Open (burnt in) and Closed (Teletext) subtitles and can be authored in upto 42 Lanuagues ..
Not I think they all were but there are say 4 nordic German French Spanish and Italian
plus English in widespread use.
is not the G0 charecter set used by Teletext see tables 35 and 36 is sect 15,6 of EN 300 706
https://www.etsi.org/deliver/etsi_i_ets/...06e01p.pdf
,,, which itself has 11 diferent national charecter set variations,
So there coudl be much opportunity for error!!!
for instance the word Sucess was output as suce
and Lesson Le on
this was due in part to the fact that the subtitle file have LC= 08 as German,,,,,
and of course ss is eszett symbol which is in the german Teletext National Charset the same position 7/E
which is the in the Englah charecter set....
..... Itwas a bit more complex as to where in the chain the substiution was done!
Because of such things the BBC Had specfications for the technical parameters of the file
which was checked at various ingest stages..
One of the first automated QC systems (C 1994 say)
And a sepcification for what the Teletext output looked like .
.Including a 4 sec auto repeat and Terminating Headers
and on UK service not used Diacrtical marks by Packet x/26 .
(which were used in the BBC Commercials channels which often had 13 Subtitle Streams
including Greek and Hebrew .. but that another story plus how itwas fitted and timed in the VBI)
key point to note - To make a Standard (ITU/ETSI/SMPTE etc) practical you need to constrain it
by specification so it meets you business needs ,
Styles -
Summerise/Paraphrase ... in the early days the Deaf community were keen on this as of course English
may not been ther first Language and thus reading speed was low.....
as subtitles became more normal the reading speed improved and the deaf commnuty wanted every word
spoken to be subtitled ,,,,,
(which the BBC trys to do all the time but there are times when they deliberately don't see later)
Sound effects - As noted Oracle was into this ... BBC feeling I think tempered by the prevailing Deaf comments
on reading speed did not want to clutter the screen.
The Colours ..... i speak for what BBC felt ....and then what happened ...
it was felt helpful to use a colour per charecter but some colours were not really useable.
so White Yellow Green and Cyan (often called Blue) were used....
The BBC had a automated syetsm for EastEnders which took teh script, parsed it tosee who was in each scene
and worked out the colours that could be used ...hence the announcments at the start...
but then the Sig Tune was shortened and there was not time to annoumce them All..
Also as time went on the visablaity of Cyan was an issue and thus White Yellow and Green were used.
(incidently the French used White for on screen and the Red & Green for Left & Right off screen dialogue)
Something you have not mentioned is
on screen postioning .....
In Teletext this soon became a bit hit an miss as the TV sets had diferent Teletext decoders with a different
number of Rows of Text e.g Fastext added a row and both the Left to rught and Up and down could differ
by a pixel or two and line or two.....
The Aim was not to obstruct the mouth or the actsion or any text ,,,
and there also was a feeeling that Left Centre or Right was helpful.....
(there are two ways of doing this - typing in the words and then settinga postion by the Justifcation code JC
in EBU3264 ... as is Left Centre Rght Or adding leading spaces to move the text to the right place.....
the use of the JC is why -if a playout sytem ignores it -all subtites are set very hard left
I will stop now about presentation - and not in this post deal with DVB subtitles,,
but as they are still derived from Teletext signal from playout to Code and Mux
(and the for Sky veiwers emitted)
there are many compromises and implemation specifications which have been lost
and some lesser channels are not even broadcasting the Right Tiresias screenfont..
Now and in the future EBU3264 is being replaced with EBU Timed Text (Or SMPTE TT for Films)
Both of which are based arround IMSC1 TT so that Streaming will use TT to the client PC or TV or phone
As will UHD probalay while legacy system will need translate/code to DVB ST or SMPTE TT for HD
and DVB ST or even Teletext fir SD (except SKy ..who use Teletext for HD as well)
Live Realtime subtitles
These are now mostly done by
respeaking - with colour change Punctuation Positioning and Cleardowns
done by mix of verbal and Pressing buttons/Clicking on screen
This obviously takes time so the subtitles are always late ...
BUT the BBC has a system which uses the fact that the Subtles take say 1/2sec to encode
while the video takes 6 secs to encode to get back say 5 secs by careful playing with the DVB PTS.
(this was explictly mentioned in the DVB ST spec back in 1997 ..... )
One thing I got done in the mid 90s was that a news reporter would sign off
"first name , second name for BBC News" which was NOT subtitled (See earlier)
which gained 2 seconds
Going back a long time
Oracle used palantype and BBC used Steno .....
But the accuracy needed was far greater than for the Courts or Hansard who could reveiw the words
as they retyped them so good Subtitle Steno operators were rare and Expensive
There still is a place for them
- and also their use of dictionaries for difficult names etc is a key to respeaking as well !!!
Just happening is some very clever AI replacing Respeaking .....
see Red Bee
https://www.redbeemedia.com/services/access-services/
they are very good at using context to resolve Homonyms ...
. different spelt words which sound the same/similar and also with dictionnaries for pasrticular contexts ,
This can also be used for prepared subtitles to get the higher quality Speech to text than normal systems
to get the text which then is just cut up or retimed (And may be colured) or tweak the output .....
I see some of the errors in this! but it is very good and quick to prepare,,,,
E.g in BBC four Coventry Cathedral - the latice spire is called a "fleche" the subtitles had "flesh"
(propably no archetectrural context yet!!!!)
Soory for being a bit long ..... and more typos than live subs
Just to correct a bit of history - ITFC (who used to prepare itv etc subtitles) was taken over by Deluxe,
a few years ago itv tendered for the service - the selected company then "got intio financail difficulties"
so itvs Access services team in Newcastle took the whole thing in house...