Microsoft Anna

From Shiarapedia
Jump to: navigation, search

Microsoft Anna Idiosyncrasies

Annachronism (n): A syllable, word, or phrase that Microsoft Anna has trouble pronouncing or inflecting. See related: Bushism.

Because Ventrilo and Discord Text-to-Speech (TTS) abuse is just too much fun. At least with Microsoft Anna - MS Sam and Toshiba Male Adult and others are meh. :) (though try 'weaponsmith' on MS Sam.) Nor has most of our group migrated to Windows 8 yet, so we have not looked for errors with the newer TTS robots yet (though try 'nini' on Microsoft Hazel).

What's the most number of butchered words we can fit in to a semi-coherent sentence that fits into the character limit of Vent TTS?

This is our list of Microsoft Anna glitches (and "science" results) and other funny text-to-speech results that we've randomly encountered so far. Also we need to split these into categories. It's not all meant to be funny, though there's certainly lots of amusing stuff.

For any comments, suggestions etc, yell in Discord if you know where we are, otherwise email me (email address is on Discussion page). Thanks for your time. :)

Installing Microsoft Anna on Windows 10

Microsoft Anna is not one of the pre-installed TTS's on Windows 10, but it is possible to get her as one of the TTS engine options on it!

To install Microsoft Anna on my Win 10 Professional build, I went to this thread and downloaded Windows 7 Text to Speech Voices Setup v1.0 installation that the user there (Butters) has so kindly compiled. It contains Microsoft Anna as well as Microsoft Lili (Chinese, and she may be broken right now).

If you are going full nostalgia, the Windows XP voices can also be downloaded, and that one contains Microsoft Sam, Microsoft Mike, and Microsoft Mary right now.

After installing this, some apps like Ventrilo will immediately have the Microsoft Anna option pop up in their TTS settings. For Discord, you will have to open the Start Menu and search for "Speech Recognition", and open the Control Panel settings for it. Go to the Text to Speech tab, and you should be able to select Microsoft Anna from the drop down menu.

Note that if you simply search for TTS or "Text To Speech", it will open up a DIFFERENT dropdown box for text to speech voices, which is used mostly for Narration (I think) and will not have a Microsoft Anna option, nor will it affect your Discord TTS voice. Search for "Speech Recognition" instead. Also do not confused "Speech Recognition (Control Panel)" with "Windows Speech Recognition (desktop app)".

Enjoy!

The Basics - wbe and the Start/Middle/End theory

wbe is a "silent" word, and is what we sometimes colloquially use to represent "wb" (welcome back) in our vent and elsewhere when people return from AFK or Internet Hiccups. Listened to by itself if everything else is silent, there is a very soft b- thump, but it's main use in this context is as a very useful catalyst to trigger other Microsoft Anna idiosyncracies.

Why? Because some words have different pronounciations at the start, middle, and end of sentences, but only when there are other words in the sentence. For example try "AK wbe AK wbe AK" and contrast that to a plain "AK AK AK". Since the wbes are silent, it tricks poor Anna into revealing a word's "start, middle, end" format without hearing spam of other actual vocalised words in between.

Consonent Sounds

Other "silent words" have also come to light beside wbe, like:

  • bhe
  • ghe
  • wfe
  • ge

But they are basically ways of making standalone consonent sounds for each letter, as they aren't actually absolutely silent when you listen to them alone, but they easily get drowned out by any audible sounds.

Some words also take these "silent words" or other syllable sounds and actually turn them into leading consonent sounds. For example,

  • bhe ake = "bake"
  • ghe ake = "gake"
  • khe ake = "cake"
  • rhe ake = "rake"
  • wbe ake = "bake"
  • wfe ake = "fake"
  • wse ake = "zake"

This also works with ike instead of ake, among other examples. We need to come up with a list of consonent sounds for funsies.

Also "din" seems to be near-silent too, but not sure why and have not done sufficient research into it.

Abbreviations

Three cheers for "op the penguin". There are definitely some really strange things that can be seen here, for example which of the states/provinces have abbreviations programmed in to Anna, and which of the elements do. Looking at the inclusions/exclusions makes you scratch your head a little. Note: We may not have tested "all" of the possible entries in each category, though for the most part I think we did.

Note a couple of them require the full stop at the end too, if it is listed with such. Capital letters vs small letters sometimes affects it too, as does whether a dot or other symbol follows the abbreviation, or if a number is required to be before or after the abbreviation, and whether the number can be numerical or spelled out, etc. Some have plurals when you tack s onto the end, some do not. Those that involve a period at the end do not ever seem to have plural forms of the abbreviation. There are a lot of 'triggers' which can decide whether an abbreviation fires or not. We've even found some oddball cases like the "iron" case below!

Elements

  • ru = Ruthenium (44)
  • dy = Dysprosium (66)
  • tm = Thulium (69)
  • lu = Lutetium (71)
  • pt = Platinum (78)

... Why do those seem to be the only ones that work? Seriously?

Iron

Yeah, iron. fe is the chemical symbol for iron (26), even though fe is not in the Elements list as 'fe' by itself does not translate to 'iron'.

But we discovered that bfe = "bee iron" ???

The following translate fe into iron (and add a letter in front):
bfe, dfe, gfe, hfe, jfe, kfe, lfe, mfe, nfe, pfe, qfe, sfe, tfe, vfe, xfe, zfe

You can capitalize any or all of the letters but one in any of the entries above and they will still work. If the entire word is capitalized however, it will not work.

The following do NOT translate fe into iron:
fe, afe, cfe, efe, ffe, ife, ofe, qufe, rfe, ufe, wfe, yfe, <number>fe, most multiple-letter combinations with fe appended at the end.

Not sure how many, if any, other examples of this kind of weirdness there are.

We've now also found that typing "Buzzfeed" translates to Anna saying "Buzzironed". So yeah, it's not only single characters with a trailing fe, it sometimes triggers with multiple leading and trailing letters too.

-ables - The little (TTS) engine that could

Another interesting quirk we've found is that some abbreviations can be directly tacked on to some letters, but will still be expanded and then appended to the front of the rest of the letters. For example, -able and -ible share this quirk, so:

  • decable = Decemberable
  • kphable = Kilometers Per Hourable (mangled like the way she says kph - see below)
  • mgrables = Managerables
  • novables = Novemberables
  • nyable = New Yorkable
  • ttsables = Text To Speechables

It doesn't work with ALL abbreviations though, for example nycable, wisable, akable and feable fail, among many others.

Months

  • apr = April
  • aug = August
  • dec = December
  • feb = February
  • jan = January
  • jul = July
  • jun = June (phonetic anyway? But they seem to share exactly the same phonetic using the wbe test)
  • mar = March (but only when a number follows it.. ie "mar 66")
  • may = May (cause it was left out alone in the cold)
  • nov = November
  • oct = October
  • sep = September
  • sept = September

States/Provinces/Cities/Countries

  • ak = Alaska
  • mh = Marshall Islands
  • ny = New York
  • nyc = New York City
  • wis = Wisconsin

... Most others don't work.

"id 37556" or something similar reads as "Idaho 3 7 5 5 6". Perhaps some other state abbreviations are that way too, working only when they precede a number?

Ordinal Numbers

  • 1st = First
  • 1st/ = One Street Slash
  • 2nd = Second
  • 3rd = Third
  • 3rd/ = Three Road Slash
  • 4th = Fourth
  • 5th = Fifth

... (all/most other numbers seem to be correct)

As you can see, punctuation directly after these breaks up the abbreviation, though it doesn't always result in two other abbreviations (2nd/ breaks up into "two en dee slash" for example).

Currency

The hallmark of these currency abbreviations is that by itself, they do not abbreviate. But when a number precedes or follows it, whether with a space or not, then it translates it into the appropriate currency. Also weird that the proper noun of the country is used in many of these cases instead of the appropriate adjective (Canada versus Canadian).

So for example, 5cad = 5 cad = cad5 = cad 5 = "5 Canada Dollars". Spelling out the number does not work for this category. Putting 1 in front uses the singular form of the currency.

You can also add decimals and it will add a subunit for each currency as well (denoted in brackets below). E.g. "2.01gbp" = "2 Great Britain Pounds and 1 Penny". As long as the decimals are at most two places over. "27.075 jpy" just becomes "27 point oh seven five japan yen".

  • aud = Australia Dollars (aud by itself is "audit" though) (decimals = cent/cents)
  • cad = Canada Dollars (decimals = cent/cents)
  • cny = China Yuan (decimals = fen/fen)
  • dm = Deutsche Marks (decimals = pfennig/pfennigs)
  • eur = Euros (decimals = cent/cents)
  • gbp = Great Britain Pounds (decimals = penny/pence)
  • hkd = Hong Kong Dollar (decimals = cent/cents)
  • jpy = Japan Yen (decimals = sen/sen)
  • usd = Dollars (decimals = cent/cents)
  • ¢ = Cents
  • $ = Dollars (decimals = cent/cents)
  • £ = Pounds (decimals = penny/pence)
  • ¥ = Yen (decimals = sen/sen)

For the currency symbols, you can additionally do "$25 million" and it will translate it to "25 million dollars". For the letter abbreviation versions, this only works if the currency comes first. ie "jpy 25 million" gives "25 million Japan yen", but "25 million jpy" gives "25 million jay pee eee".

Other

  • aaa = Triple A
  • approx = Approximate
  • apt. = Apartment
  • aud = Audit
  • ave. = Avenue
  • avg = Average
  • bbq = Barbecue (plural = Barbecues)
  • bf = Bold Face (bf2 = Bold Face 2)
  • bu = Bushels
  • bhp = Bishop
  • bldg. = Building
  • blvd = Boulevard (blvds = Boulevards)
  • bros = Brothers (bro does NOT map to Brother)
  • byu = Bayou (then again on further glance, this one may be just literal pronunciation..)
  • ch. = Chapter
  • cl = Centiliters
  • cm = Centimeters
  • co. = Company (when not at end of line)
  • conf = Conference
  • corr = Corrects
  • corrs = Correctses (she tries to pluralise 'corrects'!)
  • ct = Court
  • cu = Cubic (when not at end of line)
  • ctr = Center
  • dal = Decalitre
  • db = Decibels
  • dept = Department
  • diag = Diagram
  • div = Divided
  • dr/dr. = Doctor (when at start of line) or Drive
  • dz = Dozen
  • ed. = Education (when not at end of line)
  • esq. = Esquire
  • est. = Establish
  • ex. = Example (when not at end of line)
  • ext = Extension
  • fed = Federal
  • fl = Fluid (when not at start of line)
  • fl. = Fluid (when not the only thing in the line)
  • fn = Footnote
  • ft = Fort
  • fyi = For Your Information (but she says Information weirdly)
  • gm = Gram
  • geog = Geographic
  • gro = Gross
  • hr. = Hour (Except when end of line)
  • hrs = Hours (hr doesn't work)
  • ht = Height (hts = Heights)
  • hwy = Highway (hwys = Highways)
  • hz = Hertz
  • inc = Include
  • inc. = Incorporated
  • incl = Include
  • incl. = Included
  • inv = Inverse
  • is. = Island ("That's the way it is." = "That's the way it island")
  • jr. = Junior
  • kb = Kilobytes
  • kg = Kilograms
  • khz = Kilohertz
  • kj = Kilojoules
  • km = Kilometers
  • kph = Kilometers Per Hour (but she REALLY mangles up the per hour bit here! Wow. Perhaps puts the *pl thing below in perspective if she can mangle up hard-coded abbreviations too...)
  • kv = Kilovolts
  • kw = Kilowatts
  • kwh = Kilowatt-hour
  • l = Litre/Litres (Litre/Litres both come out sounding very different from each other..)
  • lb = Pounds
  • lg. = Large
  • ln. = Lane
  • lt = Lieutenant
  • ltd = Limited
  • mb = Megabytes
  • med = Medicine
  • mfg = Manufacturing
  • mg = Milligrams
  • mgmt = Management
  • mgr = Manager
  • mhz. = Megahertz (Gigahertz doesn't work as far as we know. The space before and the period after are required, unlike khz.)
  • mi = Mile
  • mia = Missing In Action
  • min = Minimum (If not trailed by a number and a space. "1min" = "1 minimum", "red min" = "red minimums")
  • min = Minutes (If trailed by a number and a space. "1 min" = "one minute". "two min" = "two minutes")
  • mins = Minimums
  • misc = Miscellaneous
  • ml = Milliliters
  • mm = Millimeters
  • mr = Mister
  • mrs = Missus
  • ms. = Miss (when not at end of line)
  • mt = Mount
  • mw = Megawatt
  • nat'l = National (Though many like internat'l do not translate, a few like cat'l also translate to 'Cational', bat'l goes to "Bational", rat'l goes to "Rational," and so on.)
  • ne = Northeast
  • no. = Number (when not at end of line)
  • nw = Northwest
  • op = Opus
  • perf = Perfect (perfer goes to "perfecter" but imperf doesn't go to "imperfect")
  • pg = Page
  • pk = Park
  • pkwy = Parkway
  • pl = (Pearl? Perl? Plural? I CANNOT figure out what she is saying here - you should be able to type the actual word and it should sound exactly the same.. none of those do)
  • pl. = Place
  • pon = Pontoon
  • pp = Pages
  • pt. = Point (Just pt is platinum)
  • qt = Quarts
  • rd = Road
  • rep. = Representative ("You just lost geek rep." = "You just lost geek representative")
  • req = Requirement (reqs = Requirements)
  • rps = Revolutions Per Second
  • rt = Root
  • Sat = Saturday ("Sat down on a chair" etc. Yeah I know I need to split days of the week to a sub-list of its own.)
  • se = Southeast
  • sm. = Small
  • sq = Square
  • sql = Sequel (SQL, really)
  • ssn = Social Security Number
  • Sr = Senior. Note sr, sr., SR and SR. do not work
  • st or st. = Saint (if preceded by a lowercase word and followed by an uppercase word (e.g. "test st Test") OR if at the start of a sentence)
  • st or st. = Street (all other cases, e.g. "test st test")
  • ste = Saint
  • ste. = Suite
  • ster = Sterling
  • sw = Southwest
  • str = Strength
  • tts = Text-to-Speech
  • vs = Versus
  • wk = Week
  • wt. = Weight
  • xmas = Christmas
  • yd = Yards
  • yr = Years

I find it funny MS is not mapped to Microsoft. :)

"1" or "one" in front of the units of measurements like kg, kb, km, mg, min, ml, mm, yd, yr etc turn the corresponding abbreviation into the singular form correctly. BUT ONLY if there is a space between the 1 and the abbreviation. "1kg" = "1 kilograms", "1 kg" or "one kg" = "1 kilogram".

Trailing Star

This one is a bit weird, if you add an asterisk directly after a word (with no spaces) it messes up the last syllable of the word and then adds the asterisk as a vocalised "asterisk" after. However, try "50*" versus "fifty*" and the plain "50"/"fifty". You can see that she converts "50" into "fifty", then adds the asterisk after, which mutiliates the last syllable (ty). Which is scholastically interesting insofar as figuring out Anna's "order of operations". :)

  • 50*
  • didn't*
  • safety*
  • tea

.. etc. There are many more, we just haven't really noted them down before because they're really in the category of typos, and we don't really consider typos legitimate MS Anna abuse, because it's too easy to mangle poor Anna otherwise. There is a definite pattern with asterisks though.

Symbols

Not so much errors here, but there are some amusing symbols or combinations of symbols that MS Anna does or does not verbalize. For example, ":)" has no output, but ":(" comes out as "colon" and ":P" or ":p" come out as "colon pee" etc. We should probably at some point make a list of which "colon+<key>" combinations are silent, output just "colon", or "colon" plus the trailing key (which would be the expected output).

  •  :) (Silent)
  •  :( ("colon")
  • © = Copyright
  • ® = Registered Trademark
  • alt-0177 = ± = "plus minus"
  • alt-0178 = ² = "superscript 2"

Also see the currency section above.

Degrees

° = "degrees".

Of note, a number followed by °, then a space, then C, F, or K, gives "<number> degrees celsius/fahrenheit/kelvin>" So "10° C" gives "ten degrees celsius" etc. It does not work if the number is spelled out. Decimals work. The space seems to be necessary as well.

Of further amusement, "10° C/F/K" itself makes Anna say, "10 degrees celsius slash fahrenheit slash kelvin".

She also says "degree" correctly if a 1 is placed before the ° symbol.

Onomatopoeia

Where long strings of weird characters are entered in just to hear the funny-sounding way Anna says them. Stuff like the sprinkler and helicopter noises (Dan or Jah please fill this in with the tstststst stuff and when it works or doesn't work :P).

I know there are more of these we've come across too, but I never really wrote them down as I haven't been too big a fan of them before anyway. She already mangles enough real words!

  • khe khe khe khe khe khe khe
  • lolololololololololololol
  • jah'll jah'll jah'll jah'll jah'll jah'll
  • uj
  • kan

Stuff like that. I don't like this section so will seldom add to it.

Text Filter level

Believe it or not, the text filter level (Ventrilo only?) does make a difference sometimes as well as to how badly Anna mangles something. For example, For all levels (0-3) of text filter, she says "1 5:20 am" as "one, five twenty a-meters" For text filter levels 0-1, she says "05:21:28 5:20 am" as "five hours twenty-one minutes twenty-eight seconds five-twenty aye emm" For text filter levels 2-3, she says "five twenty-one twenty eight five-twenty a-meters"

Which is just weird.. apparently it is due to interpreting XX:YY:ZZ differently in the different filter levels. In filter levels 0-1 it interprets it as a plain HH:MM:SS value, in filter levels 2-3 it interprets the XX:YY as clock time and ZZ as just a floating number, and that somehow throws off the next bit.

The proof that she interprets XX:YY as clock time can be seen by instead using the test string "12:00:00 12 am", and on filter levels 2-3 she should say "12 o-clock zero zero 12 a-meters" instead of "12 hours 0 minutes 0 seconds 12 aye emm" on filter levels 0-1.

And you can have lots of text between the "XX:YY:ZZ" and the "XX am" bit at the end, and it will still work like this - this was how it was discovered in the first place, by pasting something with timestamps and someone with filter level 3 commenting on the a-meters, and being rebutted by someone with filter level 0!


Another one that we have found is something that starts with, for example, "Awww." Eg. "Awww. Nite!" on text filter 0 is "Awww. Night!" as expected. But on filter levels 1-3, it gets mangled into "A night!" "Awww. Nini!" on text filter 0 is "Awww. Neenee!" as expected. But on filter levels 1-3, it gets mangled into "Aneenee!"

The reason for this is apparently because the text filter tries to strip out the www. bit of the line, thinking that it is part of a url, to sometimes hilarious effect.

"Unusual" Anna

For the following incomplete lists of words, Anna does something strange

  • usual company = "unusual company"
  • usual com = "unusual com"
  • usual co = "unusual co"
  • usual c = "usual c"
  • usual k = "unusual k"
  • usual cucaracha = "unusual cucaracha"

etc.

If the word following "usual" starts with a hard c/k noise, she turns the trailing 'usual' into 'unusual' instead. What? How? Why is this even possible? Who knows!

We are not sure if she does this to any other words besides usual/unusual. How weird, hey? (I bet you thought I'd say unusual there.)

Unsorted List

  • cmon
  • only logged/only lost
  • well a couple hehe
  • I might (works fine at the end of a sentence, no comma before)
  • I did
  • stand by my
  • walkthrough
  • hotkey
  • plugin
  • poo
  • gigabyte
  • lowbie
  • strat
  • autologin
  • app
  • cartoony
  • lightsaber
  • boop
  • tetris
  • webpage/webpages
  • cop/cops ("the cop" = cope, "to cop" = cop)
  • customise
  • lawn mowers
  • bribery
  • dump-ster
  • facepalm
  • wind (wind up)
  • canal
  • unqueued
  • poor destruction
  • soles
  • rotworm
  • randomly
  • refuses
  • that the earthquake (phantom "was" after "that")
  • know dammit
  • but ouch
  • pron
  • forcefield
  • kind of gnarly and
  • oddly
  • launches
  • Real time
  • helicopter (says it fine at the front of a line, but not middle or end)
  • torches though
  • dune
  • minecart
  • lagged to heck
  • literally a clone
  • groupees (groups)
  • xraying
  • though a couple were
  • 1 june 2 may 3 august
  • Lame
  • bros
  • obnoxious
  • night! night? night.
  • cred
  • pineapple (use ?, !, wbe)
  • (and) lots and lots
  • whatsoever (at the end of a line)
  • grish
  • see ya
  • some reason
  • ever
  • learned
  • iphone
  • stache
  • hundred copies/10 copies
  • in seconds
  • about par for
  • losing patience
  • and/at/that comment/s
  • but ow :( (she spells out 'oh double u' instead of saying 'ow')
  • horrid
  • per par for
  • dips
  • mature/immature
  • outlevel
  • justified homicide
  • impatience
  • pointy arrows
  • mascot
  • shock grenade
  • spoonfed
  • but outside of (or "but outside the")
  • kudos
  • scurvy
  • /spiders
  • of nostalgia
  • compulsive
  • astronomical
  • machete
  • 2 l
  • jewellery
  • chatroom
  • in allies
  • expiry
  • oddly
  • light stables, heh
  • finishes up
  • cat'l (Also try at'll)
  • if I may (she skips the 'may' entirely. Also if you try, 'when I may', she says, 'when I my' instead)
  • triple agent
  • kudos
  • Annuminas (she makes a strange sound when she says the 'nu' part)
  • mystique
  • ringtone
  • engr = "engram"?
  • grish = "gravillish?"
  • inhib
  • fiora
  • mercury
  • teehee (she pronounces it "teewhy")
  • but hahaha (vs just "hahaha")
  • reused
  • pop on in the
  • goldy-lox
  • some day technology
  • Nak is an old (or "Nak is an older")
  • whoc (sounds like "hockey")

Also, she pronounces 'hsield' and 'wsord' as 'shield' and 'sword' correctly. Very strange. She also pronounces 'wot' as 'what'.