On July 25, Volcano Engine held the 2024AI Innovation Tour in Chengdu.Bean bag large modelThe average daily usage of tokens has exceeded 500 billion, and the average daily usage of tokens per corporate customer has increased 22 times since the model was released on May 15. Zhang Xin, vice president of Volcano Engine, said that Volcano Engine is moving towards a more intelligent, industrial and regional direction, and is helping companies achieve business innovation through industry solutions, products and optimization services.
At the conference, Volcano Engine releasedBean curdLarge ModelThe latest capabilities of the Doubao AI platform include upgrades in visual images, speech synthesis, and sound reproduction. The Doubao Image Model and Doubao Text Model are outstanding in preserving the features of the original image and improving the image texture. The Doubao Speech Synthesis Model and Doubao Sound Reproduction Model have been improved in expressing emotions and restoring the characteristics of the speaker's voice.
Doubao·Tusheng Image Model: It can not only highly preserve the original image's multi-dimensional features such as character outlines, expressions, and spatial structures, but also supports more than 50 different styles, image expansion, partial redrawing, and smearing, allowing images to be creatively extended. It has been applied to applications such as Douyin, Jianying, Doubao, and Xinghui, and has served companies such as Samsung and Nubia, covering multiple fields such as mobile phone albums, tool assistants, e-commerce marketing, and advertising.
Doubao Wenshengtu Model: deeply understands information such as multiple subjects, subject-object relationships, character structure, space structure, etc., and makes picture-text matching more accurate; is better at improving the texture of the picture from three aspects: light and shadow, atmosphere color and character beauty; optimizes content with Chinese characteristics, and can provide a detailed understanding of Chinese people, objects, dynasties, geography, food, festivals, etc.
Doubao Speech Synthesis Model: It can deeply understand the storyline and characters and express emotions correctly; it retains pronunciation habits such as swallowing and accent, comparable to the real person's voice, making the voice more natural; 26 exquisite super-natural voices can meet the needs of various scenarios.
Doubao Voice Replication Model: It only takes 5 seconds to copy a high-fidelity voice, highly restores the speaker's voice characteristics and accent, supports migration across 6 major languages, and the pronunciation is closer to that of local people.
At the same time, Volcano Ark provides core plug-ins and intelligent capabilities, as well as full-cycle data security and trust solutions to help enterprises easily implement large models. The three plug-ins of Toutiao and Douyin have been upgraded, and new web page parsing plug-ins and calculator plug-ins have been added to support the diverse application needs of enterprises. Volcano Engine provides a professional version of Button, which supports low-code construction of expert "AI Bots" that fit enterprise business scenarios.
Volcano Engine has also created HiAgent, an enterprise-exclusive AI application innovation platform, to help enterprises get through the last mile of embracing big models. HiAgent fully enables the rapid implementation and continuous optimization of enterprise AI applications from three dimensions: speed, density, and thickness. Volcano Engine AI full-stack cloud relies onByteDanceIt shares a massive resource pool, supports multi-core and multi-cloud architecture, provides ultra-large-scale computing power, supports cluster networking of tens of thousands of cards, and large MoE models with trillions of parameters.