Text-to-Speech Synthesizer

Internet service “Text-to-speech synthesis” (TSS) is a high-quality tool for text processing. The system is based on free and the most widespread on the Internet scripting PHP programming language and serves to voice Belarusian or Russian  texts inputted by a user. Text-to-speech Synthesizer processes a text automatically and forms an audio file that a user can listen to, download and save to a computer.


Figure 1 – An Interface of the text-to-speech synthesizer

To get a synthesized speech a user needs to select language, insert or print a text in the input field and click “Generate synthesized speech!” (Figure 1).

For better quality of processed text, one can use the following marks:

plus / + / or aqute / ы́ / – to arrange main stress (for example, “звыча+йны“);

equal to / = / or gravis / ё̀ / – to arrange secondary stress  (for example, “тэ=леперада+ча“);

circumflex / ^ / – is used between two words to combine them into complex phonetic word (for example, “на^стале+”, “сказа+ў^бы”).

In order to download or save a generated file a user needs to click the appropriate links Listen to generated speech and download generated speech file.

While processing an electronic text to speech synthesis, the system generates additional  intermediate results. Among them there are a normalized text, phonemic text, allophonic text, etc. In order to see intermediate results of the text-to-speech synthesizer it is necessary to put a tick next to the checkbox “Show log information”.

Among other additional  intermediate results the service provides an information on each text character – its name. Like any computer-linguistic system, the synthesizer claims to use all characters that are appropriate to the settings of the system. For example, the use of a character, which does not correspond to a particular alphabetic system (Cyrillic, Latin alphabet, etc.), results in incorrect text-to-speech synthesis. In such cases it is necessary to find mismatched character and replace it with the correct one (Figure 2).

The text-to-speech synthesizer also depicts data of all the words found in a text, parts of speech, tags (contracted notations of morphological word characteristics), a list of words with stresses, placed by a user, a list of words that are absent in the system database (see Figure 2).


Figure 2 – Intermediate results of TSS: information on characters, normalized text with morphological tags, stressed tokens, unknown tokens


Figure 3 – Intermediate results of TSS: a list of homographs, intonation markers, a list of phonemic and allophonic words

Trying to determine the position of a word stress, speech synthesizer checks each word of an input text on a variation of its reading in accordance to the information provided in dictionaries, there is a search for words with the same spelling and different stresses. Information about the words that have  ambiguous stress position are displayed in the window Homographs. Also, a user receives information about intonation markers (Information markers), a list of phonemic (Phonemes) and allophonic (Allophones, Allophones information) words (Figure 3).

An additional feature of the synthesizer is transcription generator which outputs text  in 4 formats: Cyrillic, international (IPA), Simplified IPA transcription and X-SAMPA (Figure 4). More information about each of the formats can be found in the given service page “Transcription Generator“, where links to electronic resource that were used are presented.



Figure 4 – Intermediate results of TSS: four transcription formats

In addition, a user has a posibility to download files with the intermediate results of TSS, by clicking appropriate links (Downloading).


Access to the service via the API

To access the service «Text-to-Speech Synthesizer» via the API, you should send an AJAX-request (type: POST) to the address https://corpus.by/TextToSpeechSynthesizer/api.php. With an input array data the following parameters are passed:

  • text — arbitrary input text in Belarusian, Russian or English.
  • language — language of the input text. Belarusian (параметр be), Russian (параметр ru) and English (параметр en) are available.
  • voice — speech synthesizer voice:
    • for Belarusian and English following voices are available: AlesiaBelAlesiaBel (dictation mode)BorisBelBorisBel (dictation mode)BorisBelHigh;
    • for Russian following voices are available: AlesiaRusAlesiaRus (dictation mode)BorisRusBorisRus (dictation mode)BorisRusHigh.

Example of AJAX-request:

   type: “POST”,
   url: “https://corpus.by/TextToSpeechSynthesizer/api.php“,
      “text”: “Груша цвіла апошні год.”,
      “language”: “be”,
      “voice”: “BorisBel”
   success: function(msg){ },
   error: function() { }

The server returns a JSON-array with the following parameters:

  • status — processing status.
  • text — input text.
  • Tokenization results:
    • charactersText — result;
    • charactersUrl — URL, where the result is saved.
  • Text processor results:
    • tokensText — result;
    • tokensUrl — URL, where the result is saved.
  • List of tokens with accents specified by the user:
    • stressedText — list;
    • stressedUrl — URL, where the list is saved.
  • List of words that are not in the synthesizer database:
    • unknownText — list;
    • unknownUrl — URL, where the list is saved.
  • List of homographs found in the input text:
    • homographsText — list;
    • homographsUrl — URL, where the list is saved.
  • Intonational processor results:
    • prosodicText — result;
    • prosodicUrl — URL, where the result is saved.
  • Phonetic processor subtotal — phonemic text:
    • phonemicText — result;
    • phonemicUrl — URL, where the result is saved.
  • Phonetic processor subtotal — allophonic text:
    • allophonicText — result;
    • allophonicUrl — URL, where the result is saved.
  • List of allophones with characteristics:
    • allophoneCharacteristicsText — list;
    • allophoneCharacteristicsUrl — URL, where the list is saved.
  • talkingHeadCharacteristics — allophones characteristics, required for the service «Talking Head Synthesizer».
  • transcriptionCyr — transcription in cyrillic format.
  • transcriptionLat — transcription in a simplified format of the international phonetic alphabet.
  • transcriptionIPA — transcription in IPA format (International Phonetic Alphabet).
  • transcriptionXSAMPA — transcription in X-SAMPA format.
  • audio — link to the generated audio file.

For example, the following reply will be formed on the above listed AJAX-request:

      “status”: “success”,
      “text”: “Груша цвіла апошні год.”,
      “charactersText”: “Груша_груша_word_cyrillic
      “charactersUrl”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-50-16_80-94-171-2_298_be_characters.html”,
      “tokensText”: “w гру+ша_НевядомаяКатэгорыя_sbm2012initial_гру+ша_NNIFO_sbm1987_гру+ша_NFN1_noun2013_3
w цвіла+_VIIPF_sbm1987_цвіла+_дзеяслоў_verb2013_цвіла+_?_words_processed_3
w апо+шні_НевядомаяКатэгорыя_sbm2012initial_апо+шні_JJMO_sbm1987_апо+шні_JJMA_sbm1987_апо+шні_прыметнік_adjective2013_апо+шні_прыметнік_adjective2013_5
w го+д_НевядомаяКатэгорыя_sbm2012initial_го+д_NNIMO_sbm1987_го+д_NNIMA_sbm1987_го+д_NMN1_noun2013_го+д_NMA1_noun2013_5
p .
p newline”,
      “tokensUrl”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-50-16_80-94-171-2_298_be_tokens.html”,
      “stressedText”: “”,
      “stressedUrl”: “https://corpus.by/TextToSpeechSynthesizer/null”,
      “unknownText”: “”,
      “unknownUrl”: “https://corpus.by/TextToSpeechSynthesizer/null”,
      “homographsText”: “”,
      “homographsUrl”: “https://corpus.by/TextToSpeechSynthesizer/null”,
      “prosodicText”: “ГРУ+ША

      “prosodicUrl”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-50-16_80-94-171-2_298_be_intonationalText.txt”,
      “phonemicText”: “GH,R,U,+,SH,A,/,C’,V’,I,L,A,+,/,A,P,O,+,SH,N’,I,/,GH,O,+,T,/,#P4”,
      “phonemicUrl”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-50-16_80-94-171-2_298_be_phonemicText.txt”,
      “allophonicText”: “GH004,R022,U022,SH002,A323,/,C’002,V’002,I241,L002,A012,/,A221,P001,O012,SH002,N’004,I242,/,GH001,O032,T000,/,#P4,”,
      “allophonicUrl”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-50-16_80-94-171-2_298_be_allophonicText.txt”,
      “allophoneCharacteristicsText”: “GH(65ms;8000hz)

      “allophoneCharacteristicsUrl”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-50-16_80-94-171-2_298_be_allophoneCharacteristics.txt”,
      “transcriptionCyr”: “[γру́ша] [ц’в’іла́] [апо́шн’і] [γо́т]”,
      “transcriptionLat”: “[ɣrúʂa] [t͡s’v’ilá] [apóʂn’i] [ɣót]”,
      “transcriptionIPA”: “[ɣruʂa] [ʦʲvʲila] [apɔʂnʲi] [ɣɔt]”,
      “transcriptionXSAMPA”: “[Grus`a] [ʦ’v’ila] [apOs`n’i] [GOt]”,
      “audio”: “https://corpus.by/_cache/TextToSpeechSynthesizer/out/2019-09-03_17-38-19_80-94-171-2_975_be_ssrlab.wav”


Links to sources

Service Page: http://corpus.by/TextToSpeechSynthesizer/?lang=en

Калі Вы знайшлі ў тэксце памылку правапісу, калі ласка, выдзеліце гэты тэкст і націсніце Ctrl+Enter.