Home Technology peripherals AI The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

Aug 13, 2024 pm 04:34 PM
industry giant network

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

The AIxiv column is a column where this site publishes academic and technical content. In the past few years, the AIxiv column of this site has received more than 2,000 reports, covering top laboratories from major universities and companies around the world, effectively promoting academic exchanges and dissemination. If you have excellent work that you want to share, please feel free to contribute or contact us for reporting. Submission email: liyazhou@jiqizhixin.com; zhaoyunfeng@jiqizhixin.com


Since the emergence of GPT-4o in 2024, companies in the industry have invested huge resources in the research and development of TTS large models. In recent months, large Chinese speech synthesis models have sprung up, such as chattts, seedtts, cosyvoice, etc.

Although the current large-scale speech synthesis model is almost indistinguishable from real people in Chinese Mandarin, in the face of China's complicated dialects, TTS large-scale models have rarely been involved in training a unified speech synthesis of various Chinese dialects. Large models are a very challenging task.

Industry Pain Points and Technical Bottlenecks

Currently, speech synthesis large model technology has made significant progress in the field of Mandarin, but its development in the field of dialects is very slow. China has dozens of major dialects, each with unique phonetic features and grammatical structures, which makes training a large TTS model covering various dialects extremely complex.

Most of the existing large TTS models focus on Mandarin and cannot meet the diverse speech synthesis needs. In addition, the scarcity of dialect corpora and the lack of high-quality annotation data further increase the technical difficulty.

Technological innovation and breakthrough of Giant Network AI Lab

In order to solve the above problems, algorithm experts and linguists in the Giant Network AI Lab team worked together to build a comprehensive system based on the Chinese dialect system. 20 dialects, over 200,000 hours of Mandarin and dialect data sets. Through this huge data set, we trained the first large-scale TTS model that supports multiple Mandarin dialects - Bailing-TTS. Bailing-TTS can not only generate high-quality Mandarin speech, but also generate a variety of dialect speech including Henanese, Shanghainese, Cantonese, etc.

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

  • ArXiv: https://arxiv.org/pdf/2408.00284
  • Homepage: https://giantailab.github.io/bailingtts_tech_report/index.html
  • Paper title: Bailing- TTS: Chinese Dialectal Speech Synthesis Towards Human-like Spontaneous Representation

The following is the synthesis effect of Bailing-TTS Henan dialect:

Text 1:

Bianshui flows eastward Infinite spring, the Sui family palace has become dust. Pedestrians should not go up to the long embankment to look out; the wind blows and the flowers and flowers worry about killing people.

Generate voice 1:
The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly
Text 2:

I also have many hobbies. It’s nice to listen to Henan opera, and the accent is very exciting to listen to. When I have nothing to do, I can go out for a walk and take in the beautiful scenery of Henan. Fortunately, I can make some fun things, such as the braised noodles and spicy soup. Don't tell me, it's okay if I make them myself.


Generated Voice 2: The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly
Let me listen to the effect of zero-sample cloning in Mandarin:

Prompt 1: Young-Male The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothlyGenerated 1: This question, hmm, From another perspective, is it also a good thing for us? The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothlyPrompt 2: Boy-Male The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothlyPrompt 2: Hey, tomorrow is the weekend again, let’s go watch a movie together. The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothlyPrompt 3: Elderly-Female The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly
Generation 3: Speaking of our past, ah, I couldn’t finish it in three days and three nights. The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothlyPrompt 4: Toddler-Female The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothlyGeneration 4: Oh, this is what you are talking about. I picked this up when I went to the beach. The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly
We have adopted a number of innovative technologies to achieve this goal:

1. Unified dialect token specifications: We have unified the token specifications of various dialects and unified the tokens of Mandarin and various dialects There is partial overlap to provide basic pronunciation skills using Mandarin. This enables us to achieve high-quality dialect speech synthesis under limited data conditions.

2. Refined Token Alignment Technology: We propose a refined token-wise alignment technology based on large-scale multi-modal pre-training.

3. Hierarchical Mixed Expert Architecture: We design a hierarchical hybrid expert architecture for learning unified representations for multiple Chinese dialects and specific representations for each dialect.

4. Hierarchical reinforcement learning enhancement strategy: We propose a hierarchical reinforcement learning strategy to further enhance the dialect expression ability of the TTS model by combining basic training strategies and advanced training strategies.

Implementation details

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

                                                                                                                                                                                        Figure 1 Overall architecture of Bailing-TTS

1. Refined Token alignment based on large-scale multi-modal pre-training

In order to achieve refined alignment of text and voice tokens, We propose a multi-stage, multi-modal pre-training learning framework.

In the first stage, we use an unsupervised sampling strategy to conduct rough training on a large-scale data set. In the second stage, we adopt a refined sampling strategy to conduct fine-grained training on high-quality dialect datasets. This method can effectively capture the fine-grained correlation between text and speech and promote the alignment of the two modalities.

2. Based on the hierarchical mixed expert Transformer network structure

In order to train a unified TTS model suitable for multiple Chinese dialects, we designed a hierarchical mixed expert network structure and multi-stage multi-dialect tokens Learning Strategies.

First, we propose a specially designed hybrid expert architecture for learning unified representations for multiple Chinese dialects and specific representations for each dialect. Then, we inject dialect tokens into different levels of the TTS model through a fusion mechanism based on cross-attention to improve the model's multi-dialect expression capabilities.

3. Hierarchical reinforcement learning enhancement strategy

We propose a hierarchical reinforcement learning strategy to further enhance the TTS model by combining basic strategy training and advanced training strategies. Dialect expression ability. The basic training strategy supports the exploration of high-quality dialect speech expressions, and the advanced training strategy strengthens the speech characteristics of different dialects on this basis, thereby achieving high-quality speech synthesis in multiple dialects.

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

                                                           
Bailing-TTS has reached a level closer to real people in terms of robustness, generation quality and naturalness in Mandarin and multiple dialects . In Table 1 Bailing-TTS's test results in Chinese general calls and dialects

In the actual application scenario evaluation, Baling-TTS has achieved good results. In Table 2 Bailing-TTS Test results of the test result of the speaker of Chinese general calls, dialects and dialects. The multi-dialect TTS large model has been applied in many practical scenarios. For example, dubbing NPCs in games, dubbing dialects in video creation, etc. Through this technology, game and video content can be closer to regional culture, improving users’ sense of immersion and experience.

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

In the future, with the further development of end-to-end voice interaction large models, this technology will show greater potential in areas such as dialect culture protection and game AI NPC dialect interaction. In the dialect protection scenario, by supporting voice interaction in multiple dialects, the next generation can easily learn, inherit, and protect Chinese dialects, allowing Chinese dialect culture to have a long history. In the game scene, intelligent NPCs that can speak dialects and can interact with voice will further enhance the expressiveness of game content.

Giant Network AI Lab will continue to be committed to promoting the innovation and application of this technology, bringing users a smarter and more convenient voice interaction experience.

Team introduction

The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly

Makmal AI Giant telah ditubuhkan pada tahun 2022. Ia adalah aplikasi teknologi kecerdasan buatan dan institusi penyelidikan yang bergabung dengan Giant Network. Komited dalam bidang penjanaan kandungan AIGC (imej/teks/audio/video/model 3D, dsb.), merealisasikan penghasilan dan penciptaan kandungan pintar yang komprehensif, dan mempromosikan inovasi permainan. Pada masa ini, makmal itu telah membina saluran paip pengeluaran industri AI pautan penuh dalam Giant Pada masa yang sama, ia telah menyelesaikan pendaftaran model menegak besar pertama (GiantGPT) dalam industri permainan dan merupakan yang pertama dimasukkan ke dalam komersil. permohonan.

The above is the detailed content of The first large-scale TTS model that supports mixed speaking of Mandarin and dialects: Henanese, Shanghainese, and Cantonese can be spoken smoothly. For more information, please follow other related articles on the PHP Chinese website!

Statement of this Website
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn

Hot AI Tools

Undresser.AI Undress

Undresser.AI Undress

AI-powered app for creating realistic nude photos

AI Clothes Remover

AI Clothes Remover

Online AI tool for removing clothes from photos.

Undress AI Tool

Undress AI Tool

Undress images for free

Clothoff.io

Clothoff.io

AI clothes remover

Video Face Swap

Video Face Swap

Swap faces in any video effortlessly with our completely free AI face swap tool!

Hot Tools

Notepad++7.3.1

Notepad++7.3.1

Easy-to-use and free code editor

SublimeText3 Chinese version

SublimeText3 Chinese version

Chinese version, very easy to use

Zend Studio 13.0.1

Zend Studio 13.0.1

Powerful PHP integrated development environment

Dreamweaver CS6

Dreamweaver CS6

Visual web development tools

SublimeText3 Mac version

SublimeText3 Mac version

God-level code editing software (SublimeText3)

DeepMind robot plays table tennis, and its forehand and backhand slip into the air, completely defeating human beginners DeepMind robot plays table tennis, and its forehand and backhand slip into the air, completely defeating human beginners Aug 09, 2024 pm 04:01 PM

But maybe he can’t defeat the old man in the park? The Paris Olympic Games are in full swing, and table tennis has attracted much attention. At the same time, robots have also made new breakthroughs in playing table tennis. Just now, DeepMind proposed the first learning robot agent that can reach the level of human amateur players in competitive table tennis. Paper address: https://arxiv.org/pdf/2408.03906 How good is the DeepMind robot at playing table tennis? Probably on par with human amateur players: both forehand and backhand: the opponent uses a variety of playing styles, and the robot can also withstand: receiving serves with different spins: However, the intensity of the game does not seem to be as intense as the old man in the park. For robots, table tennis

The first mechanical claw! Yuanluobao appeared at the 2024 World Robot Conference and released the first chess robot that can enter the home The first mechanical claw! Yuanluobao appeared at the 2024 World Robot Conference and released the first chess robot that can enter the home Aug 21, 2024 pm 07:33 PM

On August 21, the 2024 World Robot Conference was grandly held in Beijing. SenseTime's home robot brand "Yuanluobot SenseRobot" has unveiled its entire family of products, and recently released the Yuanluobot AI chess-playing robot - Chess Professional Edition (hereinafter referred to as "Yuanluobot SenseRobot"), becoming the world's first A chess robot for the home. As the third chess-playing robot product of Yuanluobo, the new Guoxiang robot has undergone a large number of special technical upgrades and innovations in AI and engineering machinery. For the first time, it has realized the ability to pick up three-dimensional chess pieces through mechanical claws on a home robot, and perform human-machine Functions such as chess playing, everyone playing chess, notation review, etc.

Claude has become lazy too! Netizen: Learn to give yourself a holiday Claude has become lazy too! Netizen: Learn to give yourself a holiday Sep 02, 2024 pm 01:56 PM

The start of school is about to begin, and it’s not just the students who are about to start the new semester who should take care of themselves, but also the large AI models. Some time ago, Reddit was filled with netizens complaining that Claude was getting lazy. "Its level has dropped a lot, it often pauses, and even the output becomes very short. In the first week of release, it could translate a full 4-page document at once, but now it can't even output half a page!" https:// www.reddit.com/r/ClaudeAI/comments/1by8rw8/something_just_feels_wrong_with_claude_in_the/ in a post titled "Totally disappointed with Claude", full of

At the World Robot Conference, this domestic robot carrying 'the hope of future elderly care' was surrounded At the World Robot Conference, this domestic robot carrying 'the hope of future elderly care' was surrounded Aug 22, 2024 pm 10:35 PM

At the World Robot Conference being held in Beijing, the display of humanoid robots has become the absolute focus of the scene. At the Stardust Intelligent booth, the AI ​​robot assistant S1 performed three major performances of dulcimer, martial arts, and calligraphy in one exhibition area, capable of both literary and martial arts. , attracted a large number of professional audiences and media. The elegant playing on the elastic strings allows the S1 to demonstrate fine operation and absolute control with speed, strength and precision. CCTV News conducted a special report on the imitation learning and intelligent control behind "Calligraphy". Company founder Lai Jie explained that behind the silky movements, the hardware side pursues the best force control and the most human-like body indicators (speed, load) etc.), but on the AI ​​side, the real movement data of people is collected, allowing the robot to become stronger when it encounters a strong situation and learn to evolve quickly. And agile

ACL 2024 Awards Announced: One of the Best Papers on Oracle Deciphering by HuaTech, GloVe Time Test Award ACL 2024 Awards Announced: One of the Best Papers on Oracle Deciphering by HuaTech, GloVe Time Test Award Aug 15, 2024 pm 04:37 PM

At this ACL conference, contributors have gained a lot. The six-day ACL2024 is being held in Bangkok, Thailand. ACL is the top international conference in the field of computational linguistics and natural language processing. It is organized by the International Association for Computational Linguistics and is held annually. ACL has always ranked first in academic influence in the field of NLP, and it is also a CCF-A recommended conference. This year's ACL conference is the 62nd and has received more than 400 cutting-edge works in the field of NLP. Yesterday afternoon, the conference announced the best paper and other awards. This time, there are 7 Best Paper Awards (two unpublished), 1 Best Theme Paper Award, and 35 Outstanding Paper Awards. The conference also awarded 3 Resource Paper Awards (ResourceAward) and Social Impact Award (

Li Feifei's team proposed ReKep to give robots spatial intelligence and integrate GPT-4o Li Feifei's team proposed ReKep to give robots spatial intelligence and integrate GPT-4o Sep 03, 2024 pm 05:18 PM

Deep integration of vision and robot learning. When two robot hands work together smoothly to fold clothes, pour tea, and pack shoes, coupled with the 1X humanoid robot NEO that has been making headlines recently, you may have a feeling: we seem to be entering the age of robots. In fact, these silky movements are the product of advanced robotic technology + exquisite frame design + multi-modal large models. We know that useful robots often require complex and exquisite interactions with the environment, and the environment can be represented as constraints in the spatial and temporal domains. For example, if you want a robot to pour tea, the robot first needs to grasp the handle of the teapot and keep it upright without spilling the tea, then move it smoothly until the mouth of the pot is aligned with the mouth of the cup, and then tilt the teapot at a certain angle. . this

Hongmeng Smart Travel S9 and full-scenario new product launch conference, a number of blockbuster new products were released together Hongmeng Smart Travel S9 and full-scenario new product launch conference, a number of blockbuster new products were released together Aug 08, 2024 am 07:02 AM

This afternoon, Hongmeng Zhixing officially welcomed new brands and new cars. On August 6, Huawei held the Hongmeng Smart Xingxing S9 and Huawei full-scenario new product launch conference, bringing the panoramic smart flagship sedan Xiangjie S9, the new M7Pro and Huawei novaFlip, MatePad Pro 12.2 inches, the new MatePad Air, Huawei Bisheng With many new all-scenario smart products including the laser printer X1 series, FreeBuds6i, WATCHFIT3 and smart screen S5Pro, from smart travel, smart office to smart wear, Huawei continues to build a full-scenario smart ecosystem to bring consumers a smart experience of the Internet of Everything. Hongmeng Zhixing: In-depth empowerment to promote the upgrading of the smart car industry Huawei joins hands with Chinese automotive industry partners to provide

Distributed Artificial Intelligence Conference DAI 2024 Call for Papers: Agent Day, Richard Sutton, the father of reinforcement learning, will attend! Yan Shuicheng, Sergey Levine and DeepMind scientists will give keynote speeches Distributed Artificial Intelligence Conference DAI 2024 Call for Papers: Agent Day, Richard Sutton, the father of reinforcement learning, will attend! Yan Shuicheng, Sergey Levine and DeepMind scientists will give keynote speeches Aug 22, 2024 pm 08:02 PM

Conference Introduction With the rapid development of science and technology, artificial intelligence has become an important force in promoting social progress. In this era, we are fortunate to witness and participate in the innovation and application of Distributed Artificial Intelligence (DAI). Distributed artificial intelligence is an important branch of the field of artificial intelligence, which has attracted more and more attention in recent years. Agents based on large language models (LLM) have suddenly emerged. By combining the powerful language understanding and generation capabilities of large models, they have shown great potential in natural language interaction, knowledge reasoning, task planning, etc. AIAgent is taking over the big language model and has become a hot topic in the current AI circle. Au

See all articles