CandyVoice’s voice software is based on the patented technology, MUVOC®, which allows to process (control, analyze and calculate) all essential audio parameters.
These technological bricks are independent, ready to use, work in synergy and in real-time, and can be easily adapted to any product or service based on voice interaction.
MUVOC® operates at sampling frequencies of 8000 Hz to 48000 Hz, with 16 or 24 bit samples.
Muvoc® Noise Filter removes all ambient noise without altering the natural aspect of the voice, even in a complex and changing acoustic environment.
Telephone call in the street
with activation of the noise filter
during the call
Telephone call in the subway
with activation of the noise filter
during the call
Voice imitation consists of reproducing one person's voice (Target Voice) from anoter person’s voice (Source Voice), or a synthetic voice from text (Text To Speech) .
Voice Imitation requires learning the mathematical voice model of the Source and Target Voices from their respective recordings (deliberate or from archives audio files).
Then these voice models are processed in order to generate a Voice Imitation Model from the Source Voice to the Target Voice. Thus, a Source Voice can speak with the voice of the Target Voice in real time.
To obtain a better result, the recordings of the voices to be reproduced (the voices of interpreters) by the Voice Imitation, are made in the studio. Then, these recordings are pre-processed to generate the Voice Model in a semi-automatic way. At this stage, we can already hear a result that is fairly representative of the reproduction of the interpreter’s voice. Then, a supervised processing of the Voice Model will make it possible to obtain the optimal quality of the reproduction of the voice.
When the voice to reproduce comes from an archive video or audio database, CandyVoice can create its good quality Voice Model only if these archive audio recordings are good quality: are homogeneous (recorded at the same period and in the same acoustic environment), and were made in a environment without noise.
Learn more on the recording procedure
The TTS available on the market are very expensive. Their creation process is long and complex, and they offer very few voices for each language. Thanks to CandyVoice's Voice Imitation technology, all market’s TTS generic voices can be customized very quickly and easily.
Customization of Brand Voices for voice interaction-based devices, to enrich the user experience by offering a choice of a multitude voices, according to their preferences and needs.
Voice Backup, and customization of generic voices from speech aid devices (AAC, or Augmented and Alternative Communication) to the voices of patients affected by voice or speech disorders.
Customization of personal avatar’s voices with detection of mouth movements according to spoken words (Lip Synch).
TTS Voices customization allows you to listen to, rather than read, written messages (emails, text messages, social media posts, etc.) with the voice of their authors.
The option of real-time voice imitation of the game character by the player improves and enriches the player's experience, increases its immersion in the game, and also protects players data : the voice is biometric data that reveals a multitude of sensitive personal informations.
Please do not hesitate to request the technical documentation.
Voice Transformation technology can transform spoken or singing voices in a very realistic way by acting on different vocal parameters simultaneously: timbre, gender (man, woman), age (child, teenager, adult, elderly person), etc.
From just one voice, you can create a multitude of different voices. Eg: the voice of a man can be transformed into the voice of several women, children, old people or fantastic characters.
For reasons of security, confidentiality, and compliance with the GDPR, some Voice Tech players do not want to use the online voice recognition offered by GAFAM and BATX. They are then forced to create their proprietary speech recognition for their voice-based products and services. However, creating a voice corpus to train AI algorithms on which speech recognition is based represents a very significant investment. If one wishes to acquire one, unfortunately there are very few voice corpora available for sale on the market. In many cases, various players have small voice corpora which are not adapted to their needs: either too small or not rich enough.
By relying on its technological bricks, CandyVoice has developed a process which makes it possible to create from scratch, or to enrich existing voice corpora according to various specific needs: