It is generally believed in the industry that if 2023 is a "wild year" for generative artificial intelligence technology, then 2024 is expected to be the "first year" for the great development of artificial intelligence applications. With the development of AI big model and various text generation, Wensheng pictures and Wensheng video tools all the way this year, the global media industry will also usher in all-round changes in the way of content production, dissemination and consumption. The revolutionary breakthrough of technology presents the creator with a "magic pen". How should the media seize the opportunity to become a "good horse" who makes good use of this wonderful pen?
On February 26th, China’s first Wensheng video AI series was produced by the Central Radio and Television General Station.200 episodesThe animated film "Ode to a Thousand Years" was launched. The first AI mini-drama "The Myth of China" and the first AI translated English series of micro-documentaries "The Story and the Story" were also released.This is a set of concrete practices for the Headquarters to actively explore the innovative application of generative artificial intelligence technology in the media field, seize a new round of scientific and technological revolution and industrial transformation, and strive to seize the commanding heights of future industrial competition to empower the cultural industry and international communication.
With the empowerment of AI, the future of the media industry has come. How to tell Chinese cultural stories well through generative AI? How to make the new quality productivity become the biggest increment of the high quality development of the main station? Let’s start with a series of AI practices led by the reception desk.
AI big model with China aesthetic.
Let it sing, and ask for its friendly voice. The communication between people begins with the communication of language. Similarly, the communication between creators and AI has to go through the communication process of "meeting friends with literature".
In July 2023, the Headquarters and Shanghai Artificial Intelligence Laboratory released the "CCTV Listening Media Model", which was first applied to the creation of Ode to a Thousand Years.In the first season of Ode to a Thousand Years, the latest technological achievements such as controllable image generation, dynamic character generation and literary video were comprehensively used.Six well-known Tang poems, such as Ode to Goose, Crossing the Old Man’s Village and a farewell to meng haoran on his way to yangzhou of the Yellow Crane Tower, were transformed into beautiful national animation.
The program creation team uses a large number of Chinese paintings and ancient poems as training materials to communicate with the AI ? ? model. After countless collisions between artistic creation language and code language,"CCTV listening to the media model" has a unique style and strength — — China’s aesthetics.It can accurately understand the traditional cultural words such as "national style, Tang Dynasty, freehand brushwork and meticulous painting", and generate fine art materials such as meticulous painting and ink painting, so as to reproduce the artistic conception and beauty of poetry to the maximum extent.
The ratings of the first six episodes of Ode to a Thousand Years’ Poetry ranked first among all the cartoons on the satellite channel in China, reaching a total audience of 94.413 million people., the central video client’s live viewing volume reached 13.182 million person-times, and Weibo’s topic reading volume totaled 92.814 million person-times. Hundreds of mainstream media at home and abroad, including VOA, pay attention to the innovative practice of the reception desk. Many experts and scholars in the fields of science and technology, media and education expressed their support and affirmation.
Many netizens commented that "AI has shaped a new mode of production, and Ode to a Thousand Years of Poetry has opened up the imagination of the industry." "When the AI ? ? big model meets the oriental aesthetics, China’s poetry is really ‘ Live ’ Get up "and" enter the smart primary and secondary school platform early, which is convenient for teachers and students to use "… …
AI can also activate the imagination of more cultures. "Ode to a Thousand Years" has just started broadcasting. On February 29th, the artificial intelligence studio of the General Station and the Tsinghua University Yuan Universe Culture Laboratory jointly produced the "China Myth Series Mini-short Drama" and "Mending the Sky", which was released to creatively interpret the enduring code of Chinese civilization contained in myths and legends.This is the first full-process mini-drama of AI in China. Art, camera, video, dubbing and soundtrack are all completed by AI, which is another pioneering practice for the film and television application of AI technology.
The key to Wensheng video is "text". Long-term infiltration in Chinese culture, repeated training and deep learning, the research and development of "CCTV listening to media model" is becoming more and more mature, the future program production cycle will be further shortened, and the fineness of works will be continuously upgraded. More and more "understanding" AI partners can not only save time and effort in film and television animation production, game development and advertising design, but also accelerate the development process of mature application of Wensheng video, business model expansion and overseas spread.
How does AI improve the efficiency of international communication?
The application practice of AI in many international communication scenes at the main station shows that AI-assisted production can not only greatly improve the efficiency of multi-language conversion and multi-platform distribution, but also realize authentic and loving cross-cultural interaction under the escort of man-machine collaboration.
Question 1:
Suppose you are an artificial intelligence. Today’s job is to translate a documentary in English. When a dragon talks seriously about how to "fine raise dragons" and talks about "Nagada in Northeast China" with his own accent, how should you deal with it?
A: Note that to explain loong’s "ins and outs" clearly in a foreign language, you should not only translate it accurately and rigorously, but also speak it erudite and lovely.
This documentary, named "The Story", is a series of loong-themed micro-documentaries translated by the Headquarters with the full AI process empowerment, and will be launched soon.The film effectively explores the efficient application scenario of the whole process of film and television translation of AI. Using AI technology, the whole translation process from voice recognition, text translation, voice cloning to track-splitting replacement has been completed, which highly restored the style of the original Chinese film and the "character dragon design" of narration.The accuracy of translation is ensured by man-machine cooperation, which greatly improves the efficiency of translation.
In addition to the English version, the Spanish, Portuguese, Indonesian, Filipino and other versions of The Story, the multilingual version of Ode to a Thousand Years ago and other excellent programs have fully utilized the independent AI technology of the reception desk to assist in the translation. On March 10th, the English and Spanish versions of Ode to a Thousand Years’ Poetry were officially released on CGTN. At the same time, multilingual film and television translation will be included in the training plan of CCTV Listening to Media Model, which will help the construction of the national multilingual film and television translation base of the General Station and effectively improve the international communication efficiency of Chinese film and television culture.
Question 2:
There is a group of news reports that need to be translated in multiple languages and distributed overseas in the shortest time. Is there any way to quickly generate multilingual and batch video products?
A: Yes.The main station makes full use of AI technology, and realizes the geometric progression growth of news output and broadcast quantity through the form of multi-language superposition and multi-platform distribution.For example, in a short period of one week, the series of English reports of Opportunity China: A First-line Economic Observation was transformed into nearly 100 multilingual video products in 10 languages, including Spain, France, Albania, Russia, Germany and Portugal, and broadcast on overseas platforms, which strongly refuted the so-called "China economic summit theory" in the United States and the West.
These products have clear and bright pictures and real and natural sounds, which bring surprises to overseas netizens. A Russian netizen left a message: "I am also surprised that this Chinese speaks Russian without any accent. It turned out to be AI, which is really awesome!"
It is not only the reported multilingual conversion, but also the AI virtual digital avatars of several CGTN hosts. The production mode of generating video through manuscript input saves the real-life recording link, shortens the shooting and material export work that originally took up to 2 hours to only 5 minutes, and greatly improves the upper limit of single-person single-day video output.
How does the AI track run steadily and accelerate?
There is a view that the development of artificial intelligence, especially the exponential growth of generative AI, has shortened the unit of measurement of technological change from years and months to days, and may even be calculated in hours and minutes in the future. If you don’t advance, you will retreat, and if you advance slowly, you will retreat. All in or wait and see? For the media industry, it is more critical to find a way to increase productivity through technological innovation.
How does the media industry run out of "acceleration" on the AI ? ? track? In the development and application of generative AI technology, there are two main development concepts that the main station has always adhered to.
On the one hand, we should insist on original research, and strive to build an internationally leading model of Wensheng video with China cultural characteristics.Compared with Sora’s vivid visual effects, the AI animation of programs such as Ode to a Thousand Years ago has some problems, such as blurred background and insufficient coherence of characters’ actions, which shows that there is still a certain gap between the video generation model independently developed by China and the world’s advanced technology. However, at present, the Chinese corpus accounts for only 1.3% of the global large-scale model data training set, and the characteristics of China are obviously insufficient. Facts show that insisting on original research is the fundamental to promote the development of independent general artificial intelligence in China.
In 2023, the Headquarters and Shanghai Artificial Intelligence Laboratory jointly established the "China Large Model Corpus Database Alliance".On the understanding and expression of Chinese culture, and the integration with the actual production process created by the media, the AI practice led by the General Desk has unique China aesthetic and Chinese cultural characteristics.
The main station has a rich and standardized audio-visual corpus such as China Audio-visual Archive, and all materials have been tagged, which is also the largest Chinese audio-visual corpus in the world.Training with the massive high-quality audio-visual data and professional media creation knowledge of the main station, the model is expected to grow rapidly in continuous actual combat and embark on a differentiated road of China’s technological development.
On the other hand, to do a good job in platform construction, we should promote the integration of "Industry-University-Research" and continuously optimize the safe and reliable AIGC technology platform.The core of big model breakthrough lies in algorithm, computing power and data. The Headquarters will work with Shanghai Artificial Intelligence Laboratory to build an independent and controllable domestic intelligent computing platform, and further promote the newly established large-scale model research and development community and artificial intelligence studio construction. Linking the upstream and downstream of the industrial chain, the first-class scientific research institutions, universities and enterprises in the United Nations will cooperate in the fields of scientific research, demand application, safe development and industrial ecology.
Since the establishment of the main station, the whole stage has always maintained a state of being in danger, breaking through difficulties and seizing opportunities. Beginning in 2019, the head office leaders visited Shanghai Jiaotong University and other institutions to discuss the high-quality development of the artificial intelligence empowerment head office and set up relevant cooperative research institutions. After the Sora Wensheng video appeared, the leader of the main station asked the whole station to pay close attention to it and keep a close eye on it to prevent a complete defeat overnight.Artificial intelligence does not make creativity "fall from the sky", but opens up new application scenarios for "golden ideas" from all walks of life, providing innovative infrastructure and auxiliary capabilities.
After several years of practice, the Headquarters has continuously deepened the development strategy of "5G+4K/8K+AI", and at the same time vigorously promoted the innovation of media technology, transformed the new quality productivity into more refreshing quality programs. A series of practical achievements about generative AI made by the Headquarters are not a temporary follow-up "early adopters", but a series of "products made by the Headquarters" and "golden business cards" of Chinese culture. How will the AI road of the main station continue to break through the boundaries of media technology and create new imagination and fresh experience about Chinese culture for the global audience? I believe there are more scenic spots ahead.
