CereProc is excited to announce our new neural text-to-speech system, CereWave AI, powered by advanced machine learning technology. CereWave AI is available now in the CereVoice Cloud.
CereWave AI generates speech that sounds more natural than any other text-to-speech system, producing a new level of human-like emphasis and inflection. The model creates audio waveforms from scratch, using a deep neural network that has been trained using large amounts of speech. During training, the network extracts the underlying structure of the voice and learns to produce realistic speech waveforms. CereWave AI not only produces a voice that is nearly indistinguishable from human speech, but also enables full editing and control, changing it to speak any language, gender, accent or age.
Gender Change Sample:
Typical text-to-speech systems require 30 hours of recordings, but CereWave AI needs just 4 hours of data to generate a high-quality voice. The new system allows CereProc to create voices faster and more cost effectively than ever before. Our CereWave AI synthetic voices have enhanced smoothness of output, eliminating the audible joins of previous generation text-to-speech.
Using language adaptation technology, CereWave AI has already been used to enable Dr Peter Scott Morgan to speak French using his own voice, without requiring any French recordings. Our speech synthesis experts can mix data across languages, adding new languages to a voice that has never spoken it before. This exciting example displays the potential to give anyone the ability to speak any language in the world.
Peter speaking English:
Peter speaking French:
CereWave AI allows CereProc to build hundreds of voices, using a pool of existing voices and specialist adaptive techniques. This option is highly desirable for large game companies as they often require a large number of voices to be used in the development process.
Using these adaptive techniques, our team has already fused the voice of a customer suffering a degenerative illness with a relative’s voice, to produce a reconstruction of their original voice.
CereWave AI can be finely controlled while retaining naturalness and character, even allowing it to tell the perfect joke - unheard of in the text-to-speech industry!
CereWave AI has already had a profound impact on our customers, and we are proud to continue our record of innovation in the text-to-speech industry with this ground-breaking research.
As a small independent company, we are excited to announce that our English-speaking CereWave AI voices are available on the CereVoice Cloud, matching the prices and system quality of the technology giants. The possible applications for this technology are endless and we're excited to share this new evolution of speech synthesis with you.
Expressive Capability Sample: