Home>Programming related>AI Source Code
Download

Additions and corrections for Russian TTS heads from Garmin navigators

Since 2009, Garmin has been supplied by Russian TTS heads, pronouncing the names of the streets in the navigator's voice tips based on speech synthesis. Due to the complexity of the Russian language in general and the Russian toponymic names in particular, all these voices voices tips with a large number of incorrect stresses and other distortions. Pronunciation errors can be adjusted by dictionaries built into the TTS head, but in these voices these dictionaries are either empty or contain no more than a dozen names.

Most TTS heads have already been removed from official support and will not be corrected. Unfortunately, Garmin practically does not respond to comments/wishes and the latest versions of the vote, so the only thing that remains is to try to help yourself.

Fortunately, all TTS-heads .vpm files for Garmin are the same FAT file container as the .img card files and successfully understand/assemble the same Gmaptool utility in AS Subfiles mode.

And in the presence of special knowledge, it is quite possible to independently deal with the contents of the voice, replenish its dictionary and even expand its internal processing of phrases before pronouncing such useful functionality as:

  • The declination of the names of the streets depending on the context of a full phrase.
  • Recognition and division into words/fragments of the names and indices of the tracks ("turn on p23" without corrections sounds like "turn to travel twenty -third")
  • Correction of complete phrases or their fragments for a more harmonious sound ("Warning of narrow stripes" -> "Warning of the narrowing of the road")
  • Restoration of endings for numerals ("2nd" or "1st")
  • Dates processing in the names for pronouncing them in the right case ("October 25 Avenue" -> "Avenue 20 October Fifth")
  • Translation of foreign status names of roads (street, avenue, highway, etc.) . At the moment, a full translation is made for names from Estonia, Latvia, Lithuania, Hungary, France, Spain and Italy. For Finnish and German names written together with status (Schützenstraße, Lappeenranantie) , such "endings" are separated and pronounced by prepared phonemes close to the desired pronunciation.

Unfortunately, such functionality is available only for the very last generation of TTS3 votes, in the previous ones you can only replace the dictionary.

Today, there are already 3 "generations" TTS heads that are not compatible with each other- it is impossible to replace in the navigator a staged TTS-head of one generation to the vote of another generation .

Katerina 1.30 (TTS1)

The first Russian TTS head. Official support is stopped.

  • Primary: RealSpeak/Scansoft/Nuance Katerina
  • Compatible navigators: NUVI 13XX/14XX, 2455, 3790 ...
  • Voice Model: Compact (BET2 16KHZ)
  • Built -in Dictionary: there, binary format BDC (BINARY NATIVE PLATFORM DICTIONARY), maximum number of words - 10,000
  • Dictionary compiler: editor Rsude 2.1
  • Backing phrases: absent
  • Ready voice: pykuuu__katerina.vpm
  • Starting Voice Files: Pycckuu__katerina 1.30
  • Source Dictionary: Dictionary_katerina.DCT, DCT textual format in UTF-8 encoding with BOM
  • Dictionary compilation: pre -installed RSude editor (works correctly only on the Russian version of Windows XP)
  • Voice Assembly: Make_voice_katerina.bat

Milena 1.30 (TTS2)

The new Russian TTS-Got of the next generation. Official support is stopped.

  • Primary: Nuance Vocalizer Milena
  • Compatible navigators: nuvi 30/40/42/50/52
  • Voice Model: Compact (BET2 16KHZ)
  • Built -in Dictionary: There is a binary dat format
  • Dictionary compiler: Nuance Vocalizer for Automotive 5.3 + Vocon 3200 Embeded Development System 3.3 SDK
  • Backing phrases: absent
  • Ready voice: pycckuu __ilena.vpm
  • Starting Voice Files: Pycckuu__imilena 1.30
  • Source Dictionary: Dictionary_milena130.dct, text format DCT in UTF-8 encoding with BOM
  • Dictionary compilation: Build_dictionary_milena130.bat, launched a copy-copy-script dictcpl.pyc .
  • Voice Assembly: make_voice_milena130.bat

Milena 2.10 (TTS3)

The radical update of Milena of the first version (Real Directions or Natural Guidance). Official support is stopped.

  • Primary: Nuance Vocalizer Milena
  • Compatible navigators: NUVI 2x97/3x97, ..., nuvicam, the entire Drive series
  • Voice Model: Standard (DRI40_155MRF22 22KHZ) + Compact (BET2 16KHZ)
  • Built -in Dictionary: There is a binary dat format
  • Dictionary compiler: Nuance Vocalizer for Automotive 5.3 + Vocon 3200 Embeded Development System 3.3 SDK
  • Backing phrases:
    • Pre-replacement of words: ru.dct, text in the UTF-8 encoding without BOM
    • Direction/default replacement (?): DFLT.DCT, text in the UTF-8 encoding without BOM
    • Regular expressions replacements: ruleset.txt, text in the UTF-8 encoding with BOM, the first empty line is required
    • Additional regular expressions file replacements: rp_rset.txt, text in the UTF-8 encoding with BOM, the first empty line is required, performed before ruleset.txt
  • Ready voice: pycckuu __ilena.vpm
  • Starting Voice Files: PYCCKUU__Milena 2.10
  • The source dictionaries:
    • Full final dictionary: dictionary.dct, text format DCT in UTF-8 encoding with BOM
    • Black Dictionary for Russian Names: Dictionary.voc, a text format similar to DCT in the UTF-8 encoding with BOM for the editor Milan
    • Black Dictionary for the names of cities: CITIES.VOC, a text format similar to DCT in the UTF-8 encoding with BOM for the editor Milan
    • Black Dictionary for foreign names: dictionary_Foreign.voc, a text format similar to DCT in UTF-8 encoding with BOM for Milan editor
    • The final dictionary for foreign names with restored European symbols: dictionary_Foreign.dct, text format DCT in the UTF-8 encoding with BOM
  • Dictionary compilation: Build_Dictionary.bat, launched a copy of the Python script dictcpl.pyc
  • Voice Assembly: Make_voice_milena.bat

Tatyana 2.30 (TTS3)

The new TTS head, but inside the same Milena 2.10, expanded by the support of the CLC speech generation technique, instead of the “outdated” G2P, thanks to which the main fragments of phrases sound like a recorded voice of a living announcer, with a slower and even emotional pronunciation. Moreover, the CLC database is not included in the voice file, it comes separately with each update of the main “built -in” navigator card. The voice is also supplemented by several new phrases for the tips of Real Directions. Official support continues.

  • Primary: Nuance Vocalizer Milena
  • Compatible navigators: NUVI 2x97/3x97, ..., nuvicam, the entire Drive series
  • Voice Model: Standard (DRI40_155MRF22 22KHZ) + Compact (BET2 16KHZ)
  • Built -in Dictionary: There is a binary dat format
  • Dictionary compiler: Nuance Vocalizer for Automotive 5.3 + Vocon 3200 Embeded Development System 3.3 SDK
  • Backing phrases:
    • Pre-replacement of words: ru.dct, text in the UTF-8 encoding without BOM
    • Direction/default replacement (?): DFLT.DCT, text in the UTF-8 encoding without BOM
    • Regular expressions replacements: ruleset.txt, text in the UTF-8 encoding with BOM, the first empty line is required
    • Additional regular expressions file replacements: rp_rset.txt, text in the UTF-8 encoding with BOM, the first empty line is required, performed before ruleset.txt
  • Ready voice: pycckuu__tatiana.vpm
  • Starting Voice Files: Pycckuu__tatiana 2.30
  • The source dictionaries:
    • Full final dictionary: dictionary.dct, text format DCT in UTF-8 encoding with BOM
    • Black Dictionary for Russian Names: Dictionary.voc, a text format similar to DCT in the UTF-8 encoding with BOM for the editor Milan
    • Black Dictionary for the names of cities: CITIES.VOC, a text format similar to DCT in the UTF-8 encoding with BOM for the editor Milan
    • Black Dictionary for foreign names: dictionary_Foreign.voc, a text format similar to DCT in UTF-8 encoding with BOM for Milan editor
    • The final dictionary for foreign names with restored European symbols: dictionary_Foreign.dct, text format DCT in the UTF-8 encoding with BOM
  • Dictionary compilation: Build_Dictionary.bat, launched a copy of the Python script dictcpl.pyc
  • Voice Assembly: Make_voice_tatiana.bat

Dictionary of cities

The names of settlements are made in a separate dictionary and are included in the final dictionaries of Milena of both versions and Tatyana. They are also part of the additional dictionary for Milen's voice from the Nuance Vocalizer Expressive 2 TTS For Android assembly from 4PDA.TO to use Mapcamdroid, which can voice the warning of cameras and so on, including the names of settlements, through TTS.

  • Ready dictionary: userdct_ruur_milena.dat
  • Auxiliary regular pronunciation correction expressions: rex_rur.dat, text in the UTF-8 encoding with BOM, the first empty line is required
  • The source dictionaries:
    • Full final dictionary: alerts.dct, text format DCT in UTF-8 encoding with BOM
    • Black Dictionary for Correction of some notification of Mapcamdroid: alerts.voc, a text format similar to DCT in UTF-8 encoding with BOM for Milan editor
    • Black Dictionary for the names of cities: CITIES.VOC, a text format similar to DCT in the UTF-8 encoding with BOM for the editor Milan
  • Dictionary compilation: Build_Dictionary_Aleerts.bat, launching a copy of the Python-script dictcpl.pyc

The finished dictionary and, if desired, the regular expressions file are copied to the voice data on the Android device with the name VocalizerEx2 according to the installation instructions.

Links

  • Recent versions of updated votes
  • Garmin TTS-headed forum
  • Discussion of the phonetics of names built into the HERE cards used in Garmin
  • Forum for using TTS heads in Android on 6pda.to

Notes

  • Dictionaries, including case forms of names, and regular expressions, are the same for Milena of the second generation and Tatyana
  • The finished dictionary of cities is updated with each change in the starting dictionaries and is not included in the votes for Garmin
  • To view links to the forum https://4pda.to/forum/ We need preliminary registration
Expand
Additional Information
Related Applications
Recommended for You
Related Information All