Google to Update Its Speech Service

Cloud Speech-to-Text and Text-to-Speech services received new interesting features
30 August 2018   855

The Google Cloud team announced the stable release of the Cloud Text-to-Speech speech synthesis API with experimental audio profiles function and support for several new languages. Service for decoding audio Cloud Speech-to-Text has learned to recognize different speakers and independently determine the language of the recording from several possible ones.

Along with the transition to a stable working regime, the API for translating written speech into spoken language now supports a number of new languages ​​and voices created with the help of WaveNet technology. In total, 14 languages ​​and dialects are available, which are spoken by 30 standard "voices" and 26 ones that are based on WaveNet.

Audio profile function is available in beta mode. It allows you to automatically optimize the audio file for a particular device: smart watches and other wearable gadgets, smartphones, headphones, conventional and stereo speakers, smart home audio systems, car speakers. You can also set the mode to "default".

Cloud Speech-to-Text API received the function of recognizing speakers by voice. Using machine learning, the system, when transcribing, separates the replicas of different people and marks them with numbers. However, at the beginning of the audio file processing, you need to specify the number of speakers.

Also, the Google Cloud team added the automatic language detection function to the record. Using the API for their applications, the developer can specify up to 4 languages ​​in one query. At the time of writing, the tool supports 120 languages.

Telephone Filter
Telephone Filter

The technology of speech synthesis Google used for a long time only in its own products. For third-party developers it became available in March 2018 with a choice of 32 voices and 12 languages. And the service of decoding the oral speech used to be called the Cloud Speech API, and the current name was received in April 2018, along with new models for analyzing calls and video.

Nvidia to Open SPADE Source Code

SPADE machine learning system creates realistic landscapes based on rough human sketches
15 April 2019   679

NVIDIA has released the source code for the SPADE machine learning system (GauGAN), which allows for the synthesis of realistic landscapes based on rough sketches, as well as training models associated with the project. The system was demonstrated in March at the GTC 2019 conference, but the code was published only yesterday. The developments are open under the non-free license CC BY-NC-SA 4.0 (Creative Commons Attribution-NonCommercial-ShareAlike 4.0), allowing use only for non-commercial purposes. The code is written in Python using the PyTorch framework.

Sketches are drawn up in the form of a segmented map that determines the placement of exemplary objects on the scene. The nature of the generated objects is set using color labels. For example, a blue fill turns into sky, blue into water, dark green into trees, light green into grass, light brown into stones, dark brown into mountains, gray into snow, a brown line into a road, and a blue line into the river. Additionally, based on the choice of reference images, the overall style of the composition and the time of day are determined. The proposed tool for creating virtual worlds can be useful to a wide range of specialists, from architects and urban planners to game developers and landscape designers.

Objects are synthesized by a generative-adversarial neural network (GAN), which, based on a schematic segmented map, creates realistic images by borrowing parts from a model previously trained on several million photographs. In contrast to the previously developed systems of image synthesis, the proposed method is based on the use of adaptive spatial transformation followed by transformation based on machine learning. Processing a segmented map instead of semantic markup allows you to achieve an exact match of the result and control the style.

To achieve realism, two competing neural networks are used: the generator and the discriminator (Discriminator). The generator generates images based on mixing elements of real photos, and the discriminator identifies possible deviations from real images. As a result, a feedback is formed, on the basis of which the generator begins to assemble more and more qualitative samples, until the discriminator ceases to distinguish them from the real ones.