D
STEP 1: RECORD, RECORD, RECORD
First, we choose a voice actor with a great sounding voice who is fluent in a certain language. Then we bring him or her in to talk with us – for hours and hours and hours.
We record the voice actor saying a range of speech units, from whole sentences to syllables. These can be recipes, sports results, magazine articles or anything that lets us capture the natural sound of the actor’s voice. This covers examples of all the possible sounds in a given language.
STEP 2 AND 3: SORTING Antonyms THE SPEECH UNITS AND BUILDING A VOICE DATABASE
Antonyms Now that we have thousands of recorded sound files, we need to sort them out and organize them. The speech units are labeled and segmented by phones, syllables, morphemes, words, phrases, and sentences.
These speech units are used to build a large voice database. The voice is now ready for you to use.
STEP 4: SO, LET’S CREATE SOME AUDIO FILES Antonyms
You sit down at your computer and open one of NeoSpeech’s products. You type in the text you want transformed to speech.
STEP 5: NATURAL LANGUAGE PROCESSING
From the language processing end, your text is normalized and broken down into phonetic sounds before going through a series of analyses to understand the structure of the sentences as well as to determine the context of the word for pronunciation. This is called Natural Language Processing.
Through these processes, we are able to establish prosody—rhythm, stress, and intonation—and produce natural sounding speech.
STEP 6: CHOOSING THE RIGHT SPEECH UNITS
This is where the Natural Language Processing (NLP) Part and the Voice Database come together to start producing speech.
Once the NLP is complete, our software searches the voice database and chooses the speech units that best fit together to produce the sounds associated with your text. This is called Unit Selection (hence the name, Unit Selection Synthesis).
Do'stlaringiz bilan baham: |
