Following the September version update, Guangcone Intelligence learned from Tang Jiayu’s circle of friends, co-founder and CEO of Shengshu Technology, that the Vidu large model will be upgraded again this week, and the Vidu-1.5 version will be launched soon.
The update direction of this version still focuses on extending the generalization ability and subject consistency of large models. The previous version focused on the consistency of a single subject, while the latest version can understand and integrate multiple concepts such as characters, objects, and environments, and follow user instructions to generate relevant video results of multiple subject fusion within 30 seconds, taking the lead in video creation. Multi-agent consistent generation.
In addition to Vidu, since September this year, according to incomplete statistics, mainstream AI video generation platforms including Bytedance’s Jimeng AI, Kuaishou Keling AI, Runway, Zhipu Qingying, Aishi Technology PixVerse, and pika have all A version update has been performed.
Currently, in the hot AI video generation track, large model start-ups and major Internet companies have entered the game. After intensive product launches in the early stage, it has now entered the stage of product iteration and upgrade competition.
Through the updated content of each version, it is not difficult to find that the general direction of iteration of AI video generation large model capabilities is still the duration of the generated video, the stability and continuity of the picture, and the consistency of the subject before and after.
But at the same time, various players began to "divide" in actual functional applications, each with its own emphasis. Some small and medium-sized players also began to find their own market segments.
For example, the latest version of Runway has updated Act-One, which can accurately reproduce the facial expressions of real people to AI characters, to enable 3D AI camera control. PixVerse has launched various Halloween special effects, venom special effects, etc.
Regarding this round of updates to various AI video generation platforms, Chen Kun, founder of Xingxian Culture and producer of the AI original fantasy IP "Mountains and Seas Mirror", believes: "The biggest update should be the expression migration of Act-One, which provides a better way for character performances. Basic possibilities." As for the consistency and stability of the characters, "there is progress, but there is no intergenerational progress."
According to Vicky, the creator of AI video, compared with the original product in the first half of the year, the latest updated AI video platform has not only iterated on the underlying model capabilities, but also updated its functions, such as head and tail stitching, image quality and Frame rate supplementation, dubbing and other functions, "the improvement of these functions is actually more comprehensive than in the first half of the year."
If the first half of 2024 is an arms race on the AI video generation track, then the second half of the year will be a small-step version update cycle.
At this stage, the competition between Byte and Kuaishou is still fierce. Small and medium-sized manufacturers are beginning to find their own unique tracks, and some companies are focusing on overseas markets, and have also achieved the effect of "flowering domestically and fragrant outside the wall".
Undoubtedly, the fighting at this stage may seem mild, but it has a substantial impact on the platform's own positioning and future development direction, as well as the subsequent sustainable growth of user groups and quantity.
"Jimeng is a little behind." This is an objective evaluation given by users of the AI video generation platform.
As one of the first batch of AI video generation platforms last year and a product of Byte, Dream AI’s video generation effects have been criticized by users, and are being beaten by players such as Runway and Pika.
In June this year, Kuaishou, Byte's direct competitor in the field of short videos, officially launched the "Keling" large video generation model on its official website and quickly emerged from the industry. At the same time, more and more AI video generation platforms are springing up, and the AI video generation track is completely booming.
Under strong competitive pressure, Bytedance, as the first echelon of domestic AI products, has made up for the shortcomings of video generation, which has become a top priority, and its speed to catch up is beyond imagination.
On September 24, the 2024 Volcano Engine AI Innovation Tour was held in Shenzhen. Chen Xinran, the former head of Douyin Art, appeared as the head of Jimeng AI and Cutting Market and Operations, and announced that Jimeng AI has been connected to Doubao’s latest Video generation model.
At the same time, ByteDance released two video generation models, Seaweed and Pixeldance, of the Doubao model family, and invited small-scale testing to creators and corporate customers through Jimeng AI and Huoshan Engine respectively.
On November 8, Dream AI, an AI content platform owned by ByteDance, announced that Seaweed, a video generation model developed by ByteDance, is officially open to platform users.
According to ByteDance, the beanbag video generation model Seaweed that is open for use this time is the standard version of this model. It only takes 60 seconds to generate a high-quality AI video of 5 seconds, which is 3 to 5 minutes ahead of all domestic industry standards. Requires generation time.
Jimeng AI also revealed that the Pro versions of two video generation models, Seaweed and Pixeldance, will also be available for use in the near future. The Pro version model can realize natural and coherent multi-shot actions and complex interactions with multiple subjects, and overcomes the consistency problem of multi-shot switching. It can maintain the consistency of the subject, style, and atmosphere when switching lenses, and is suitable for movies, TV, and computers. , mobile phones and other devices.
ByteDouyin and Kuaishou, as the leaders of domestic short video platforms, their competition has shifted from short video and e-commerce to the field of AI. Objectively speaking, Douyin is ahead of Kuaishou in all aspects. But only in the field of AI, Kuaishou has given a beautiful counterattack.
Since its instant success in June, Kuaishou Keling has actually had several iterations of smaller versions.
But in terms of the underlying large model capabilities, on September 20 this year, Kuaishou released version 1.5 of Keling, which is connected to a new generation of models and has achieved significant improvements in image quality and dynamic quality. The original model has also added a new function - motion. Brush, the generation effect is more controllable.
"Keling 1.5 is very strong. It can be said to be the most realistic among all models. Compared with Runway, it has basically overcome the previous problem of character deformation." AIGC entrepreneur AIgen (stage name) said to Lightcone Intelligence.
In the actual generated video effects, comparing Keling and Runway, we can see that with the same prompt word, both have a strong effect on the stability of the actual character subject, but the video effects generated by Keling can automatically unlock the face. expression.
"Runway can actually generate facial expressions on its own, but the effect is very weird." said Yamjiang AIgen. However, the abilities of Keling AI and Runway are random and not fixed.
In fact, it can be seen that Keling AI and Runway are superior in actual generation effects, and in terms of understanding prompt words, Keling AI is indeed at the forefront, but in the future it will still need to be continuously iteratively upgraded to be able to This ability is solidified.
(Runway, prompt word: a female model wearing new Chinese clothing, showing off her style, with colorful smoke floating in the background, provided by AIgen)
(Keling AI, prompt word: A female model wearing new Chinese clothing is showing off her style, with colorful smoke floating in the background, provided by Yamjiang AIgen)
However, after Jimeng launched the latest video to generate a large model, Vicky believes that its model capabilities and UI design are not much different from Keling. At the same time, during the internal testing of the Jimeng Platform Pro version model, it can easily control the movement range and actions of the screen.
As the leading short video platforms in China, Kuaishou and Bytedance have laid out their AI video generation tracks. The ultimate goal is to attract and retain users’ attention, which requires continuous production of novel, high-quality, and creative products. content.
Based on this, AI short dramas have also become one of the focuses of competition between Bytedance and Kuaishou Keling.
In July this year, the AI short drama "Mountains and Seas and Strange Mirrors: Chopping Waves" created by "Keling AI" attracted widespread attention. The short drama became the first AIGC original fantasy micro short drama in China.
In September, Kuaishou Xingmang Short Drama and "Keling AI" launched the "Xing You Lingxi-AI Short Drama Creation Competition". It is reported that the competition encourages more people to join the creation of AI short dramas through various measures such as traffic rewards, honorary awards, and content signing.
Byte is also not to be outdone. While Dream AI is teaming up with Bona Pictures to release the first AIGC-generated science fiction short drama "Sanxingdui: Future Apocalypse", it is also teaming up with many "super creators" on the Douyin platform to achieve co-creation, inviting There are high-quality fans and high-influence experts on the platform who have jointly joined the "Super Creator Alliance" program, hoping to build the largest virtual creation community in China.
But at this stage, whether it is Douyin or Kuaishou, the content created by film and television creators on their video platforms is “difficult to break out of the circle.” Vicky said, “Because the entire market has not yet been formed, and C-end users do not know how to use it. What is it here for? There will be some commercial demand for the head, but there is not much demand, and the overall situation is not stable.”
After all, there are still relatively few professional creators in the world at this stage, and AI video generation large model technology is still in its early stages.
Therefore, as the leading video platform, the competition between Byte and Kuaishou is becoming increasingly fierce. In addition to the battle for underlying AI technology and products, what is more important is who can take the lead in exploring the path of technology-enabled content. After all, if the platform can gather more innovative content creators, it can create a community ecosystem that is more concerned and loved by users.
Of course, in addition to Byte and Kuaishou, other players in the AI video generation track have also begun to "divide". Some small and medium-sized manufacturers have also begun to explore and find their own path to differentiated competition.
On short video platforms such as Douyin and Kuaishou, the content created by some creators may be difficult to break out of the circle, but some videos containing ghost and animal special effects are extremely popular, such as the AI-generated video of He Jiong and Huang Lei suddenly fighting. .
For players in the AI video generation track, ByteDance and Kuaishou are competing in a full range of technology and content ecosystems, while other small and medium-sized players are more focused on segmenting the track and identifying their own platforms and Product positioning has become the basis for survival and development.
At the end of October, Runway’s CEO made it clear in an open letter that Runway is not an AI company, but a media and entertainment company. “I think the era of AI companies is over.”
Based on this, while major companies are competing to improve the length, fidelity, and smoothness of AI video generation, Runway has clearly developed its own characteristics in the AI video track - making AI that specifically serves art, media, and entertainment.
Judging from Runway's actual video generation effects, its effects on character stability and consistency can be said to be at the forefront. In addition to basic technical capabilities, in the latest version update, the two new functions launched by Runway, although small, will provide great convenience and huge cost savings to animators, game developers and filmmakers.
Runway can be said to be one of the most popular products among film and television practitioners. In addition to its technical strength, the most important thing is its cost-effectiveness.
"Runway is so fragrant. We use Keling sparingly, but Runway is unlimited. It doesn't matter if you smoke it hundreds of times a day." AIgen said, "The randomness of AI videos is still very strong. If you charge per-view, it may be difficult for ordinary creators to afford this cost.”
On the other hand, if you use 1,000 yuan to buy points, you can buy 15,000 keling points. Each time you use 35 keling points, 1,000 yuan can only be generated 428 times. For real entrepreneurs, it is basically not enough. “Judging from the frequency of more than 200 videos I generate every day at Runway, the points purchased by Keling for 1,000 yuan are basically burned out in 2 days.” said AIgen. .
In the previous article of Guangcone Intelligence, "The explosive AI videos, big manufacturers go to the left, start-ups go to the right", it was also mentioned that the membership charging method adopted by each platform at this stage cannot be commercialized. For closed-loop entrepreneurs, the subsequent payment rate and willingness to pay will not be very high. Nowadays, it seems that even for entrepreneurs who can achieve a closed-loop commercialization, cost-effectiveness is also a key factor affecting their use of products.
In addition to Runway, Pika and Pixverse have also found their own tracks. It can be seen from their latest updated version that the focus of these two companies is to train some special effects that users can directly use. "Although the metaphor may not be appropriate, it is a bit close to the stickers made by Douyin before." Potato Jiang AIgen said.
For example, during the Halloween period at the end of October, the Pixverse V3 version added many new Halloween-themed special effects, including zombie mode, wizard hats, monster invasion and other themed effects, as well as AI pinch special effects similar to Pika’s popular AI pinching and video extension functions. Users can add an additional 5-8 seconds of content to existing videos, and can precisely control the content direction of the newly added clips.
With the recent release of the "Venom: The Last Dance" movie, PixVerse has launched a new special effect "We Are Venom" video effect based on the latest video model PixVerse V3, which can generate cool venom animations from pictures with one click.
Currently, this kind of ghostly special effects is very popular among users on social platforms. Previously, Pika launched the AI pinching special effect in version 1.5. Once launched, it was loved by users. It also relied on this wave effect to achieve overtaking in corners. Conch AI, which started growing around the same time as Pika, actually relied on character performances and meme expression packs to directly ignite overseas public opinion and overtake others in a corner.
Pika’s AI pinching effects
In fact, although Conch AI was launched late, industry practitioners have a high opinion of Conch AI. "Conch AI's performance in character movements is very good. The recent AI-generated video of He Jiong and Huang Lei fighting was generated by Conch AI," said Yangtaojiang AIgen.
However, more importantly, Conch AI has achieved the effect of "flowering domestically and fragrant outside the wall". As an AI video generation platform launched overseas by domestic AI company MiniMax, search popularity continued to rise once it was launched.
According to statistics from the "AI Product List", the number of visits to the Conch AI web version soared 860% in September, ranking first in global and domestic growth in September. Overseas users have shared their experience on social platforms, and it is generally believed that Conch AI is one of the best AI video generation tools currently on the market.
With the popularity of its products in overseas markets, MiniMax has been at the forefront of the large model Six Little Tigers in terms of commercialization capabilities.
In comparison, platforms such as Vidu and Zhipu Qingying are constantly evolving in terms of subject consistency, character stability, and video generation duration, but they have not yet formed their own style and uniqueness. competitive advantage.
Although AI video generation technology is constantly evolving and has derived unique segmented tracks. However, the Cinda Securities research report also shows that AI video generation technology still needs to be further improved in terms of character consistency, required duration, and picture quality to meet commercialization standards.
At the same time, the current mainstream AI video tools are still in the stage of competition for video generation, and most of them are single-function products. It still requires a variety of different video creation tools to be used in series to achieve the effect of directly outputting commercializable videos.
In the future, the AI video generation large model platform will still need to continue iterative evolution.