The first rule I added was to indicate that "ur" at the end of a word is stressed so that it now can pronounce words like Gravur, Tastatur, Klausur, Zensur, etc. correctly. The second rule I added was a word which it pronounced incorrectly before. The change to the existing rules were so that it pronounces kanal and signal everywhere with the stress on the a.
Those are more rules for correct pronunciation. The ey rule at end of word was removed to make espeak pronounce names like Meyer, Speyer, etc. correctly which it didn't do before. The other rules I added are also for correct compound word pronunciation. German's got a lot of compound words, a lot of them espeak can't say, but adding every single word there exists would be hard I guess, those are at least a few
I hope that will be my last change. Most of them, like Zitat, gedulds, etc have been added for correct compound word pronunciation. The A: phoneme in komma and beta sounds more natural for me; I also did Komma this way because it matches the way espeak reads it when written as an actual word.
These are some improvements to the german voices, both default and mbrola. It fixes some sounds of phonemes and also pronunciation of words. I also modified the mbrola phoneme and voice files to make them sound more natural. The stressLength setting is that it matches the way espeak deals with the lengths.
Don't omit <marker> event for words which are recognized as pairs (such as "of the").
Lang=Spanish, fix missing "y" in numbers such as "21".
Lang=EN-US, Don't add 'intrusive r' between phoneme [i@] and a following vowel.
git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@268 d46cf337-b52f-0410-862d-fd96e6ae7743
Amplitude changes now apply also to mbrola voices.
"voicing" attribute is now used in mbrola voice files to set the voice loudness.
SSML commands such as <mark> and <prosody> were ignored if they were inside a multiword group (defined in *_list). This is now fixed so that the command is processed at the end of the group.
git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@267 d46cf337-b52f-0410-862d-fd96e6ae7743
-x command-line option. Fixed possible buffer overflow when writing phoneme output.
Slavic languages, changed behaviour of [v] during voiced/unvoiced propagation.
When looking for a word boundary when breaking long clauses, allow for greater word size (Southern Indian languages have big words 3-byte UTF-8 character codes).
$combine. Don't combine with the nerxt word if it is translated using a different language.
git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@262 d46cf337-b52f-0410-862d-fd96e6ae7743
Don't speak punctuation names faster at high speeds.
lang=ml, recognise "virama + U+200d" (zero-width joiner) to indicte 'chillu' characters.
Fix problem of punctuation character announcement repeated at the start of the next sentence after speaking is stopped.
Fixed crash due to large numbers of box-drawing characters in text input.
Fixed phoneme translation of 't' sounds for US-English voices.
git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@261 d46cf337-b52f-0410-862d-fd96e6ae7743
Karl Eick