商汤Seko产品发布会
文章语言:
简
繁
EN
Share
Minutes
原文
会议摘要
Commercial Soup Technology's Chief Product Officer Zibin introduced the upgrade of SQO 2.0. Since the release of SQO 1.0 in July 2023, it has attracted 200,000 creators, with 30% focusing on short film series and 20% dedicated to slow drama. Addressing the pain points of creation, SQO 2.0 introduces multiple gathering functions for content creators, optimizes the UI interface, supports the creation of long texts and multiple drama series, and enhances the quality and production efficiency of the drama series. Strengthening the compatibility with domestic chips, reducing costs, promoting the combination of AI and the short drama market, and promoting the production of high-quality content. Commercial Soup has collaborated with Xuebao Studios to release AI short dramas, signed a contract with the Yangtze Film Group to jointly incubate AIGC works, and unleash creativity and imagination.
会议速览
Shang Tang SQO2.0 conference focuses on AI creation, introduces product upgrade process, and discusses the relationship between AI creators and the industry. Product upgrades include UI improvements, the first launch of multi-aggregation power capability, solving the problem of content consistency, supporting short and slow drama modes, and meeting the needs of creators.
Multi-agents support long text multi-episode creation, providing functions such as outline analysis, character and scene editing, to achieve consistency and plot control. The addition of a plus feature facilitates extended creation, while single-shot canvas editing improves efficiency, unified asset management, and reduces cross-software operations.
Discussed how AI technology can lower the barrier to content creation, allowing individual creators to produce high-quality works at a low cost, indicating a trend of disruptive changes in the film and television industry from low-end to high-end. Emphasized the importance of balancing the output of two cultures, "yangchun bai xue" and "Xialiba people," as well as the role of AI platforms like SQO in providing virtual studio functions, enabling individual creators to independently complete large productions that previously required team collaboration.
The discussion on how AI technology can reduce the cost of science fiction content creation, from millions of dollars to tens of thousands, and achieve efficient output. The Suzhou certified AI picture-in-picture studio has become a model for OPC companies, promoting the founding of OPC by cycle users, unleashing creative potential. By using positive and negative prompting words and reference image technology, the difficult problem of subject consistency and diversity is solved, and the quality of content during the multi-aggregation process is improved.
Discussed the importance of consistency and efficiency in video production through the collaboration of models and agents, especially in applications involving multiple people speaking and storyboard management. Emphasized the importance of the 30-degree principle in filming, as well as the crucial role of models in the development of video generation technology. Showcased the innovation and practical applications of multi-person speaking technology, and pointed out the importance of integrating models more effectively into workflows to enhance accuracy in speech synchronization.
Discussed the application of AI technology in the short drama industry, especially how to reduce costs and improve content quality through techniques such as model distillation, promoting the development of the short drama market towards a high-quality direction. It pointed out the landmark event of the Chinese short drama market surpassing the professional film market, as well as the huge market opportunities brought by the combination of AI and short dramas. Emphasized the importance of cost control for tool applications, as well as the strategy of achieving a balance between efficiency and cost through innovative technology.
The dialogue emphasized the important role of domestically-produced chips in optimizing video generation models, as well as the application cases of AI tools in inspiring student creativity in the field of education. Both sides also announced the launch of CQL 2.0, aiming to provide discounts to more users, while also jointly organizing a contest with the storytelling community to encourage creators to participate.
要点回答
Q:What important findings were discovered among the group of creators after the release of CQO 1.0?
A:After the release of CQO 1.0, we found that over 200,000 creators quickly gathered. Among them, 30% of the creators focus on creating short films and series content, while 20% of the creators specialize in making long-form dramas. This phenomenon prompted us to communicate deeply with creators, understanding the issues and pain points they encountered when using the AIGC model, especially the difficulty in maintaining consistency in characters and scenes in content creation.
Q:What upgrades have been made in addressing pain points of creators in CQO 2.0?
A:In CQO 2.0, we have launched the ability to generate multiple aggregates and opened short drama and slow drama modes to meet the production needs of more creators. Specifically, we have completely upgraded the UI of SQL 1.0, providing an immersive interface and interactive prompts that make it easier for creators to create. Through the multiple aggregation function, creators can create long-text multi-episode dramas, achieve precise control over character and scene shaping and editing throughout the entire series, maintain consistency between characters and scenes across multiple series, and also have overall control over the plot of the series.
Q:How does CQO 2.0 improve the interactive experience of episode design and storyboarding?
A:In CQO 2.0, each image supports interactive modification, and the canvas has been upgraded to allow for detailed interactive processing. This means that creators can complete most of the modification work within one software, without the need to switch between multiple tools or search through cloud documents, greatly increasing work efficiency. In addition, episodic design allows for consistency in story, characters, and scenes when creating multiple episodes, and enables fine control over key framing and overall generation of the entire series.
Q:How does AI technology drive the development of the film and television industry, especially for low-cost production?
A:The application of AI technology has enabled the production of short and slow dramas, which were originally low-cost, to achieve nearly the quality of movies. This heralds a disruptive change in the entire industry in the future. AI can integrate high-end production processes that require a large amount of manpower and cost into larger operational platforms in a low-cost and efficient manner, promoting the development of the film and television industry towards higher-end audiences and greater creative capabilities. At the same time, the development of the film and television industry should also take into account the sinking market of both high art and popular preferences, complementing each other and jointly promoting the prosperity and development of the film and television industry.
Q:In traditional film and television production, why is it difficult for individual creators to obtain enough resources to direct large-scale projects?
A:Due to resource limitations and age issues, young creators often do not have the opportunity to direct large-scale works and obtain enough filming resources, especially the resources required for traditional film and television production that ordinary individuals find it difficult to afford. The SQL platform provides a virtual studio and office, simulating the selection of roles, props, costumes, and scene layout through AI and agent technology, allowing creators to complete outputs that previously required millions of yuan with just tens of thousands of yuan.
Q:How can the SQL platform help creators reduce costs and improve efficiency in their work?
A:On the SQL platform, with the support of AI models, creators can complete short films in a few hours that would have taken a week or even several days in the past, significantly reducing production costs and increasing efficiency. For example, some sci-fi movies or animated films have been completed by actors and crews using this platform.
Q:How was the AI PIP Studio in Suzhou officially certified as the first OPC company?
A:The official certification of AI painting company Cycle in Suzhou as the first OPC company in the area is due to its excellent representation as a Cycle user, demonstrating efficient creative abilities and outstanding results on the platform. This marks the official and social recognition of AI creative companies in unleashing creative possibilities.
Q:How can AI models ensure consistency and diversity of subjects in multi-clustered scenarios?
A:In AI models, in order to maintain subject consistency, in addition to the forward reference graph, a reverse reference graph will also be constructed. The goal is to generate results that are similar to the subject but not completely identical, in order to achieve both subject consistency and responsiveness to cue words. At the same time, AI models can also consider the consistency of multiple individuals and multiple gathering scenarios, achieving natural responses and coordination in complex scenarios such as multiple people talking simultaneously through the construction of efficient reference graphs and the use of relevant technologies.
Q:How to apply the "30-degree rule" in AI filming to meet the viewing habits of the audience?
A:The "30 degree rule" refers to the distance between multiple cameras when shooting, which is typically kept at 30 degrees or more to avoid a jarring effect during editing, ensuring a smooth transition and meeting the viewing habits of audiences over the past 130 years.
Q:Can we further embed AI models into the current workflow to improve the quality of speech-to-text matching?
A:Current AI models have already achieved significant effects in lip syncing, but embedding them better into existing workflows so that all participants can effectively utilize this tool, thus improving the quality of lip syncing in the film, is a direction worth exploring.
Q:In the application of AI technology, what is currently the most worth paying attention to market?
A:The most noteworthy market currently is the micro-drama sector in China, which holds a significant share in the international market, and with the combination of AI and short films, market opportunities are further expanding. This industry is shifting from low-quality series to high-quality boutique productions, with many high-quality dramas improving their picture and overall quality with the help of AI capabilities.
Q:How will the CQ2.0 tool perform in terms of cost?
A:The CQ2.0 tool demonstrates significant cost advantages and is one of the tools with the lowest cost per frame in short-distance races. Through efficient compression technology and the design of batch generation of frames, the cost of generating a large number of frames can be greatly reduced. At the same time, through distillation technology to optimize model structure, it not only maintains the inference cost unchanged but also improves the overall model ceiling, effectively reducing the learning difficulty and cost of each model.
Q:What work have you done in the process of adapting domestic chips?
A:During the localization process of domestic chips, we have optimized the model extensively and specifically adapted and optimized it for domestic computing power. For example, we have open-sourced a tool called Light x2V, which supports the deployment of video generation models and image generation models on domestic chips, enabling video generation models to run nearly real-time on domestic chips. We are about to launch a large number of domestic inference chips and gradually integrate CQL products into the domestic ecosystem. Users are welcome to participate and promote the use of solutions that combine domestic chips with CQL.
Q:What are the applications and impact of CQL in the field of education?
A:CQL has been successfully applied in the field of education, such as a Chinese language teacher in a school in Shanxi using CQL to transform students' essay models into videos, enhancing students' sense of identity and creativity. In the future, CQL will launch a collaboration school program to allow more teachers and students to experience AI technology, unleash creative imagination.
At the same time, CQL's website slogan also emphasizes the idea of helping those with ideas to realize creative expression and complete their works, and will accompany promotional activities and co-organized contests with the storytelling community, encouraging creators to participate in submissions.

SENSETIME-WR
Follow





