{"id":103848,"date":"2022-04-22T15:36:52","date_gmt":"2022-04-22T13:36:52","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/novyj-vypusk-sistemy-sinteza-rechi-silero"},"modified":"2022-04-22T15:36:52","modified_gmt":"2022-04-22T13:36:52","slug":"novyj-vypusk-sistemy-sinteza-rechi-silero","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/novyj-vypusk-sistemy-sinteza-rechi-silero","title":{"rendered":"New Release of the Silero Speech Synthesis System","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
A new public release of the neural network-based Silero Text-to-Speech synthesis system is now available. The project primarily aims to create a modern high-quality speech synthesis system that rivals commercial solutions from corporations and is accessible to all without the need for expensive server hardware. <\/p>\n
The models are distributed under the GNU AGPL license, but the developing company does not disclose the training mechanism for the models. To run it, PyTorch and frameworks supporting the ONNX format can be used. The speech synthesis in Silero is based on the use of deep modified modern neural network algorithms and methods of digital signal processing. <\/p>\n
It is noted that the main issue with modern neural network solutions for speech synthesis is that they are often only available as paid cloud solutions, while public products have high hardware requirements, lower quality, or are not finished and ready-to-use products. For example, to run one of the new popular end-to-end synthesis architectures, VITS, seamlessly in synthesis mode (i.e., not for training models), graphics cards with more than 16 gigabytes of VRAM are required. <\/p>\n
Despite the prevailing trend, Silero solutions successfully launch even on a single-threaded Intel x86 processor with AVX2 instructions. On a quad-thread processor, synthesis allows for generating between 30 to 60 seconds of audio per second at 8 kHz, about 15-20 seconds at 24 kHz, and approximately 10 seconds at 48 kHz. <\/p>\n
Key features of the new Silero release: <\/p>\n
\n
The model size has been reduced by half to 50 megabytes;\n
Models can make pauses;\n
Four high-quality voices are available in Russian (and an infinite number of random voices). Pronunciation examples;\n
Models have become 10 times faster and can synthesize up to 20 seconds of audio per second on 4 processor threads in 24 kHz mode;\n
All voice options for one language are packed into a single model;\n
Models can accept entire paragraphs of text as input, with support for SSML tags;\n
Synthesis operates immediately at three selectable sampling frequencies \u2014 8, 24, and 48 kilohertz;\n
The 'childhood problems' have been resolved: instability and word skipping;\n
Flags have been added for controlling the automatic placement of stress marks and for including the letter '\u0451'. <\/ul>\n
Currently, there are 4 voices in Russian publicly available for the latest synthesis version, but the next version with the following changes will be released soon: <\/p>\n
\n
The synthesis speed will increase by another 2-4 times;\n
Synthesis models for the CIS languages will be updated: Kalmyk, Tatar, Uzbek, and Ukrainian;\n
Models for European languages will be added;\n
Models for Indian languages will be added;\n
Models for the English language will be added. <\/ul>\n
Some of the systemic issues inherent in Silero synthesis include: <\/p>\n
\n
Unlike more traditional synthesis solutions such as RHVoice, Silero synthesis does not have integration with SAPI, easy-to-install clients, and integrations for Windows and Android;\n
The speed, while unprecedentedly high for such a solution, may be insufficient for real-time synthesis on weak processors at high quality;\n
The automatic accent placement solution does not handle homographs (words like \u0437\u0430\u043c\u043e\u043a and \u0437\u0430\u043c\u041e\u043a) and still makes errors, but this shortcoming will be addressed in future releases;\n
The current synthesis version does not work on processors without AVX2 instructions (or requires special modification of PyTorch settings), as one of the modules inside the model is quantized;\n
The current version of the synthesis essentially has only one dependency, PyTorch; everything is 'packed' within the model and JIT packages. The source code for models is not published, nor is the code for running models from PyTorch clients in other languages;\n
Libtorch, available for mobile platforms, is much bulkier than the ONNX runtime, but the ONNX version of the model is not yet provided. <\/ul>\n