AI video generation and rise Stable Diffusion Wanxing Technology and other intensive "show muscles"

At the time of the "anniversary" of AIGC's popularity, OpenAI kept "fighting in the palace" here, while "the rise of the scarlet ball" was on the other side. In just half a month, from Meta Adobe At Runway, Stable Diffusion and Wanxing Technology, technology giants at home and abroad and AI Unicorn companies are accelerating into the AI Wensheng video track.

On November 16, Meta launched two AI Video editing New features. The first tool is called "Emu Video". It can automatically generate a four second video based on subtitles, images, text descriptions, etc; Another tool is called "Emu Edit", which allows users to modify or edit videos more easily through text commands. This tool provides users with a way to delete or replace objects and people from pictures without any professional image editing experience.

On November 22, Stability AI released its latest AI model, Stable Video Diffusion, which can generate videos from existing pictures. It is reported that the Stable Video Diffusion actually consists of two models - SVD and SVD-XT. SVD can convert still pictures into 14 frames of video, while SVD-XT can increase the number of frames to 24 frames.

On the same day, Runway, an "old" enterprise in the field of AI video generation, also officially announced its new "motion brush" function. Users only need to paint an area or subject in the picture, select a direction for it, and add motion intensity. The motion brush can add controlled motion to the user's generation.

Runway's motion brush function can generate controlled motion

On November 23, Adobe confirmed the acquisition of Rephrase.ai, an AI startup. The latter can be combined with text scripts and user avatars to convert into videos with user images, which can be used in marketing, customer communication, holiday greetings and other business scenarios. It is worth noting that this is Adobe's first acquisition in the AI field.

On the evening of November 24, AIGC Software A-share listed company Wanxing Technology (300624. SZ) launched a short video "A Girl's Life" created by AI on its official video number, demonstrating its video generation ability of multimedia model. In the large model drive The user can control the content of the start frame of the video simply by entering the story outline or clip script, and realize the complete generation of new video content through the large model capability.

Screenshot of the video Girl's Life generated by Wanxing's "Sky Curtain" model

Previously, Wanxing Technology had predicted the "sky curtain" model. It is reported that "Skyscreen" is the first multimedia model focusing on video creative application in China. It is composed of video model, audio model, picture model and language model, and has the ability to make films with one key AI art design, student music, audio enhancement, sound effect analysis, multi language dialogue and other core capabilities, and continue to iterate on multimodal key capabilities such as vision and hearing.

As the leader in the AIGC field in China, Wanxing Technology is continuously deepening the AIGC layout, and has launched new AIGC creative products such as Wanxing Love paintings, Wanxing broadcast, Wanxing intellectual performance, Wondershare Kwicut, Wondershare VirtuLook, etc; Wanxing Meow Shadow Wondershare PDFelement、 Wanxing PDF AIGC capabilities have been integrated into products such as, Yitu Pictorial, Yitu Brain Pictorial, and Mockitt, an overseas version of the Ink Knife, and they have been continuously optimized and updated.

On November 29, the AI startup Pika Labs officially launched the 1.0 version of the AI video generator. Pika Labs 1.0 adopts a new AI model, which can generate and edit videos in different styles such as 3D animation, animation, cartoon and film. Once the product was launched, it attracted numerous AI leaders to "call", and quickly gained $55 million in financing, becoming the latest "top stream" of AI video generation.

The latest AI video generates "top stream" Pika Labs

In fact, in this round of AI wave, Wen Sheng Wen and Wen Sheng Tu have been developing in parallel. ChatGPT represents the first breakthrough in text generation. Midjournal has pushed Wen Sheng Tu to be available to everyone, and with the gradual extension of the timeline, countless new players have been welcomed. Under the guidance of the "pioneers", the new players obviously took fewer detours, and the competition between the two tracks has become increasingly intense.

The underlying principle of video is the combination of multiple frames of images, which is different from the direct generation of literary texts and literary maps. Literary videos/graphic videos need to add a time dimension on the basis of literary maps. This means that computing costs are high - a short video contains about 30 images per second, and a single video clip has hundreds of thousands of frames. To ensure the consistency of space and time between each frame, a large number of computing resources are required.

In addition, complex information is difficult to process - video has visual dynamic information, and modeling video content becomes very challenging after adding time information between different frames.

Finally, the high intensity requirement for algorithm stability - the high requirement for video coherence means that the Wensheng video/Pictographic video algorithm needs to ensure that the AI generation of each frame needs to be highly relevant, continuous and smooth.

Everything is one and two sides. AI video generation technology is still in rapid iteration and rapid evolution, and how many new opportunities will emerge from it is still unknown. The only thing that is certain is that the players who have entered the game now have obviously taken the lead in competing for the best.

This article is an original article. If it is reproduced, please indicate the source: AI video generation and rise Stable Diffusion Wanxing Technology and other intensive "show muscles" https://news.zol.com.cn/844/8445255.html