Welcome to the WeChat subscription account of “Sina Technology”: techsina
Text/Xiaoyang
Source: Connection Insight (ID: lxinsight)
AI drawing has become “hot”.
In 2022, a large number of AI drawing tools will flood into the market, and the software represented by Stable Diffusion will be open sourced, which will greatly reduce the development threshold of AI drawing applications, and further push the concept of AI drawing to a new climax.
On mainstream short video platforms and social platforms, a large amount of AI drawing content is gradually flooding. Related topics on Sina Weibo have a total of nearly 30 million views. On the Douyin platform, the content related to “AI drawing” has exceeded 250 million views. For content creators, after experiencing UGC and PGC, an era belonging to AICG has gradually arrived.
The lowering of the development threshold and the explosion of the C-end have gradually attracted the attention of capital. Overseas players such as Jasper and Stability AI have repeatedly obtained financing, and their valuations have increased significantly. They have also begun to explore commercialization models such as payment and subscription. Although the domestic market has not yet had such a boom, the rapid development of overseas players has also provided a reference for the domestic market.
Big factories are not willing to miss this opportunity. Since this year alone, major Internet companies such as Baidu, Tencent, ByteDance, and Meitu have joined them. Baidu has launched an independent product, Wenxin Yige, and other major companies have added related functions to their mainstream apps. QQ has launched an AI drawing applet, and Douyin has also added related special effects.
The concept of AI drawing can bring traffic to big factories in the short term, and even increase the stock price, but if you want to occupy a place here, you need a lot of technical investment, including high hardware costs and years of training costs. Let AI generate more diverse and more sophisticated content to establish barriers to competition.
Dachang eyeing the next traffic pool
In 2022, a large number of AI painting tools will be launched, and there will be many domestic players. Products made by domestic start-up companies such as TIAMAT, Pirates of Dreams, and Yijian AI have aroused enthusiastic responses from the market.
In August of this year, Baidu released the AI painting platform Wenxin Yige based on the technologies of Flying Paddle and Wenxin Large Model. In December of this year, Wenxin Yige invited the famous Shanghai-style painter Le Zhenwen to complete the unfinished draft based on the talented woman Lu Xiaoman in the Republic of China, and released the painting of the same name “Unfinished To be continued”, which was sold at an auction for 1.1 million yuan .
At the end of the year, Tencent, ByteDance, Meitu and other major companies have joined the boom, and they have all added AI painting functions to their mainstream software.
Tencent launched the “QQ Small World AI Painter” activity, and launched the product “Me in Different Dimensions” based on the Stable Diffusion model. Users can open and use it in the QQ applet, and upload photos to generate a two-dimensional image.
“Me in Different Dimensions” did not attract much attention in the domestic market, but unexpectedly gained attention overseas, which once made the platform’s servers unable to bear it. In the related topic of Twitter “DifferentDimensionMe”, many users are trying out and writing tutorials, and Tencent has also added error messages in English.
QQ small world interface
A similar story also happened on Meitu Xiuxiu, which recently launched an AI painting function. According to Qimai data, as of December 5, with the newly launched AI painting function, Meitu continued to occupy the first place in the free list of the App store in Japan, and at the same time, ranked top in the App store rankings in 13 countries or regions. ten.
Douyin has also launched AI painting special effects. You can enter the special effects by searching for keywords in the app, and you can generate paintings without setting prompt words.
It is worth noting that more AI drawing-related functions have appeared on Tiktok. For example, in the “artificial intelligence green screen” function, after the user enters text, AI can generate images of different styles on the white background. Perhaps in the future, Douyin will also add a series of related functions.
In addition, the AI painting functions of Douyin, Meitu Xiuxiu, and QQ are built in the APP, focusing on convenience, without the need to input keywords or parameters, further reducing the threshold for image generation.
However, at the same time, the problem of homogenization has also arisen, and more cases of overturning have been born, such as a warm wedding photo mixed with a “third party”, a photo of the owner and the cat turned into a hug of two people, etc.
In a short period of time, these car rollover cases have gained the attention of users in a funny and entertaining way, and attracted a lot of traffic for the product.
In order to continue to attract users, avoid user loss, and increase user payment rates, “big manufacturers” also need to improve the maturity of AI painting products, and iterate more functions for AI painting.
Why does AI painting attract big manufacturers?
In the past two years, the concept of AIGC has attracted a lot of capital, and star companies have been born.
Before the popularity of AI drawing, AIGC concepts such as AI collaboration and video were already well known. AIGC refers to the production method of automatically generating content through AI. The content produced includes text, images, audio, video, etc., and is considered to be a technical scene directly related to VR, Metaverse, etc.
Overseas, emerging companies such as Jasper, Stability AI, and OpenAI have gained great attention.
In October this year, Stability AI announced that it had received US$101 million in financing, with a post-investment valuation of US$1 billion. Its most well-known product is the AI painting tool Stable Diffusion; in the same month, Jasper, an AIGC start-up company that focuses on text generation, also announced that it had received 1.25 100 million U.S. dollars in Series A financing, with a valuation of 1.5 billion U.S. dollars; this year, OpenAI, which was established earlier, was also revealed to be negotiating with Microsoft to obtain new financing. The current OpenAI valuation is close to 20 billion U.S. dollars.
As the popularity of related concepts continues, major domestic manufacturers have entered the game one after another, and they can also get a wave of traffic in the short term, labeling themselves as AI paintings. The stock price of Meitu has soared after the launch of the Meitu app.
But in the long run, the potential market value of AI graphics can bring new growth points to major manufacturers, which is why they rush to the battlefield one after another, hoping to occupy a place.
Overseas players such as Stability AI have been able to increase their valuations in a short period of time because the continuous iteration of AI technology in recent years has greatly increased the charm of AI graphics.
The work that best represents the highest level of current AI graphics is the “Space Opera House” created by a game designer Jason Allen through Midjourney. Art category winner.
“Space Opera House”, source wiki commons
Compared with the influence brought by the awards, the appreciation of this painting allows people to see that AI painting is not only able to produce light and low-quality works, but also has the possibility of entering the palace of art.
Keen capital markets, breakthroughs in new technologies, and the potential for AI-generated content all represent unprecedented opportunities. The development potential of the AI drawing market is huge. According to Guotai Junan Research Report, the penetration rate of AI painting in the field of image content generation will reach 10%-30% in the next five years, and the market size will exceed 60 billion yuan.
Regarding whether players can realize commercial realization based on this technology, AI drawing tools have begun to try pay-as-you-go and subscription systems.
Generating a high-definition large image on Stable Diffusion requires an average of 5-10 virtual coins (10 euros equals 1000 virtual coins), and users can pay according to the number of pictures required.
Midjourney adopts a subscription system and provides two subscription packages: $10 per month, providing up to 200 pictures, and extra charges; paying $30 per month, providing unlimited picture generation services.
Midjourney’s payment information, the source of which is its official website
This paid service and subscription service is mainly for content creators. At present, many companies are beginning to adopt the content form of AI painting. AI works are characterized by short time consumption and low cost, which can effectively reduce enterprise costs.
Previously, Liu Junyong, CEO of a game company called Guyu Games, said in an interview with the media that under financial pressure, Guyu Games streamlined the art team, replaced the art content in the game with AI painting, and reduced the original art cost of 1 million yuan to 100 yuan. Yuan.
In the fields of advertising production, IP creation, film and television production, among designers, architects, editors and other people, AI graphics can be used.
The overseas “Atlantic Monthly” used AI painting instead of external procurement. In China, the People’s Daily, with the technical support of Baidu, launched an AI painting version of “My Motherland” MV. In the future, the increase in demand will also increase the scale of AI painting creators.
In addition to content creators, AI drawing players can also develop monetization channels by providing orders to companies and governments.
The imagination of the AI painting tool lies in the fact that it can expand a variety of content generation services, including text-to-image conversion, AI audio and video generation, AI animation generation, etc. The AIGC market has huge potential, and players can find more growth points.
In 2022, when the epidemic continues, AI drawing has quietly become a new track for major manufacturers to push, but at present, it is still unknown whether major manufacturers can tap the commercial value they expect in this highly imaginative outlet.
The layout of major manufacturers is cautious, and the profit model of AI painting needs to be solved
In the layout of the AI drawing field, the big manufacturers are not radical, they are cautious, and they prefer to embrace new technologies with lightweight tools.
The prudent attitude of major manufacturers is not surprising. Although domestic and foreign players are already exploring commercialization models, they still need to break through many obstacles in order to make profits.
The cost of AIGC product development is not low, and the high hardware and training costs have raised the industry threshold.
The young Stability AI now maintains more than 4,000 Nvidia A100 GPUs running on the Amazon cloud. According to Business Insider, Stability AI has more than $50 million in operational and cloud computing spending.
In terms of training costs, a large proportion of labor is required in the initial stage of modeling. To generate high-definition images, more training time and cost are required. Algorithm optimization and data accumulation will continue for several years, and then machines will provide manual services. With the increase of user scale, players also need to increase the calculation speed per unit time in order to serve more people.
Compared with high development costs, users are currently less willing to pay. Domestic players such as Pirates of Dreams and Yijian are obtaining traffic for free. They allow users to accumulate points by signing in and watching videos, and the points can be exchanged for free image generation opportunities.
In the initial stage of the industry, players will change the market at a low price for a period of time, but in the field of software tools, users’ awareness of payment is already weak, and AI drawing is a new thing, and players need to go through more difficult market education.
At the same time, there are also many chaos in AI painting. On Weibo, Xiaohongshu and other platforms where a large number of painters gather, many people said: AI painting is “splicing materials” and infringes the copyright of the original author. Overseas, the American media CNN reported that many artists accused Stable Diffusion of using their works but destroying their jobs.
Deep learning is the algorithmic basis of AI drawing. It collects existing images, classifies and recognizes them, and finally AI can draw.
Those high-quality works that are out of the circle are only individual phenomena, such as the award-winning AI painting “Space Opera House”, which was completed after being fine-tuned by the author more than 900 times after it was generated by AI.
At present, the algorithm models of some platforms are still prone to the effects of “splicing” and “plagiarism”. However, there is still room for AI drawing to improve content generation capabilities. At the same time, copyright issues involved in the training library can be avoided by cooperating with creators, purchasing copyrights or sharing revenue. A review team can also be established to protect creators. rights and interests.
In addition, sensitive information such as pornography, violence, terror, and gore is also a big mess, which increases the operational risk of the platform.
Meitu Xiuxiu mentioned that the platform can conduct audits to avoid irregular behaviors of users. During the user’s use, image input and output will carry out content understanding. At this stage, the algorithm will filter sensitive information such as pornography, violence, horror, and blood based on the tags generated by content understanding. Secondly, in the stage of generating AI painting results, the generated results will be reviewed again.
AI painting is still a new thing, the legal definition is not clear, and it becomes more difficult for creators to protect their rights. The industry will still have chaos of “stitching” and “plagiarism” for a certain period of time. At the same time, irregular behaviors of users also need to The platform has established a sound review mechanism. B-end companies may not be able to ignore the risks involved, and look at this outlet more cautiously.
At present, the development logic and profit model exploration of domestic and foreign players are in progress. Perhaps in the near future, AI painting can take off in a real sense, usher in a large number of paying users, find a way of compliance, and attract more The joining of B-end enterprises.
However, for large manufacturers, the market potential of AI graphics attracts them, but their business layout has not yet reached a strategic level, and their resource investment is not too large. Perhaps, they are also waiting for the moment when the industry explodes.
(Disclaimer: This article only represents the author’s point of view, not the position of Sina.com.)
This article is transferred from: https://finance.sina.com.cn/tech/csj/2022-12-15/doc-imxwttix6542150.shtml
This site is only for collection, and the copyright belongs to the original author.