Voices from Amazon’s Alexa, Google Assistant and other AI assistants Far away Old school GPS devices, but they still lack the rhythm, vowel and other qualities that make speech good, humane. NVIDIA has unveiled new research and tools that can help you capture those natural speech qualities by training AI systems with your own voice, the company announced. Interspeech 2021 Conference
To improve AI voice synthesis, NVIDIA’s text-to-speech research team has created a model called RAD-TTS, a winning entry in the NAB Broadcast Convention competition for the development of the most realistic avatar. The system allows an individual to train text-to-speech models with their own voice, including pacing, tonality, timbre, and more.
Another RAD-TTS feature is voice conversion, which allows a user to deliver a speaker’s voice using another person’s voice. That interface gives subtle, frame-level control over the pitch, duration, and power of a synthesized voice.
Using this technology, NVIDIA researchers have created more conversational-sounding voice descriptions for themselves. I have AI video series Synthetic use instead of human voices. The purpose was to get a description that matched the tone and style of the video, which has not been well done in many AI described videos to date. The results are still somewhat robotic, but I’m better than any AI description I’ve ever heard.
“Through this interface, our video producers could read and record video scripts themselves, and then use AI models to convert her speech into a female narrator’s voice. Done, “NVIDIA wrote.
NVIDIA is distributing parts of this research-optimized to run efficiently on the NVIDIA GPU, of course যে anyone would want to try for GPU-accelerated conversation AI via the open source NVIDIA NeMo Python toolkit, which is available at the company’s NGC hub content and other content. .
“Several models are trained on the NVIDIA DGX system with thousands of hours of audio data. Developers can fine-tune any model they use, speeding up training using mixed-perfect computing in the NVIDIA Tensor Core GPU,” the company wrote.
All products offered by Engadget are selected by our editorial team, independent of our parent company. Some of our stories have affiliate links. If you buy something through one of these links, we can earn an affiliate commission.