Kreativität, Aufnahme und Nachbearbeitung: Abdeckung des gesamten Prozesses. Das KI-Film- und Fernsehgroßmodell von Orca Entertainment wurde auf der Yunqi Conference vorgestellt.
"The series of AI film and television large models jointly developed by Orca Entertainment and Tongyi Wanxiang has covered the entire process of 'pre - production creativity - mid - production shooting - post - production', helping the creative team to launch higher - quality content." On September 25th, at the AIGC Innovation and Development Forum of the Yunqi Conference held in Hangzhou, Jing Yi, the person in charge of the AI film and television industrialization algorithm of Orca Entertainment Group, delivered a speech titled "Exploration of the Application of Large Models in the Film and Television Industry", and for the first time introduced the current technological progress of Orca Entertainment in the AI film and television large models to the public.
It is understood that this set of AI film and television large models of Orca Entertainment includes breakthroughs in technologies such as the camera movement model CameraWeaver, wearable - free AI facial motion capture technology, and AI facial expression enhancement. It can help the film and television creative team quickly implement storyboard concepts in the creative stage, efficiently generate 3D scenes during shooting, and automatically simulate and optimize micro - expressions in the post - production of animation.
The series of AI film and television large models of Orca Entertainment has covered the entire process of "pre - production creativity - mid - production shooting - post - production"
In the content creative stage of the film and television industry, the creative team usually needs to use a combination of language description and hand - drawn sketches to present the envisioned scene pictures, which often takes a lot of time. "Although there are already several text - to - image models on the market that can generate relatively detailed pictures, they still cannot meet the requirements in terms of shot types, composition perspectives, etc., which the creative team pays more attention to." Jing Yi said. In response to this problem, the "film - grade text - to - image" large model developed by Orca Entertainment, different from the general models on the market that pursue "visual divergence", has made targeted adjustments in the precise control required by the film - grade, accurately solving the industry pain point of the creative team's "difficulty in implementing creative pictures".
"We hope that the model can accurately execute the text instructions of the creative team, so we have set strict standards for the model in terms of shot types, composition, perspectives, spatial relationships, etc., so as to accurately restore the creative intention, form a visual reference for the creative team, and greatly improve the efficiency of pre - production communication and storyboarding." Jing Yi said.
The "film - grade text - to - image" large model developed by Orca Entertainment
Jing Yi introduced that in the mid - production shooting stage, Orca Entertainment has developed a technology model for image - to - 3D scene generation, which can support the rapid generation of multi - camera storyboard reference pictures. If the creative team wants to view the storyboard effects from other perspectives, there is no need to redraw 2D storyboards or build 3D scenes. The model can reconstruct the 3D space based on a single picture, synchronously generate multi - perspective pictures, and also support camera movements such as panning, tilting, and zooming at any angle, helping the creative team quickly confirm the spatial layout.
Compared with the UE virtual scenes that take several weeks to build by traditional methods, this technology model can generate high - precision 3D spatial scenes within 2 minutes, achieving a hundreds - fold increase in efficiency. Moreover, there is zero error in spatial perspective, meeting the strict standards of the film and television industry. Currently, it has been applied to the collaborative work of storyboarding and virtual shooting pre - rehearsal in internal projects. Jing Yi revealed that during the previous internal testing, the score of this model on WorldScore had ranked first on the list.
In the post - production aspect, Orca Entertainment has also launched an innovative solution of wearable - free facial capture + AI fine - carving for animation, which makes up for the defect that facial devices cannot capture micro - expressions. This solution can accurately restore and complete the micro - expressions and mouth shape details of actors through a self - developed facial expression generation model. "We have upgraded the work that might have taken animators several weeks to manually key - frame (manually adjust key frames by animators) in the past to automatically generate micro - expressions through the facial expression generation model, and the quality has reached the film - grade standard." Jing Yi introduced. With this model, the efficiency of manual key - frame adjustment has been greatly improved, and this technology was successfully applied to projects in the first half of this year, becoming a key tool to improve the accuracy of micro - expressions and production efficiency.
The innovative solution of wearable - free facial capture + AI fine - carving of Orca Entertainment
"AI does not replace creators, but liberates them." Jing Yi said. "The goal of Orca Entertainment is to allow creators to free themselves from cumbersome technologies and focus on the story itself. Just as calculators have not replaced accountants, AI will not replace directors, but will become a good helper for creators to jointly create higher - quality content for the audience."