How will big models transform the content creation industry?
In the PC era, editors mainly used professional software to produce videos; in the App era, non-professionals also began to use mobile tools to create videos and create self-media; in the AI era, with the explosion of generative AI technology, the industry will have What are the new changes?
An AI video creation tool called "Dujia" owned by Baidu attempts to answer these questions with a series of innovative functions.
On DuPlus, creators can generate a video with rich content by inputting a copy.
▲Case study "Appreciating the sunset on the Chao Phraya River in Thailand"
It supports one-click AI polishing of copywriting, titles, and video materials, helping creators save a lot of time in editing and modification.
▲AI polishes video titles and intelligently recommends replacements
What if you don’t even want to write a script? Users can use Dujia’s AI rewriting function or directly generate copywriting scripts based on hot news. Not only can the required video be generated, but the source of the information can also be accurately traced.
▲Write video script copy based on hot recommendations
Still want an anchor to show up? Nearly ten AI digital humans with different styles are available to choose from, and realistic anchor announcements can be generated in one minute.
▲Case "Is it easy to drive an electric car at minus 34 degrees Celsius?" "exhibit
The AI-generated video track is accelerating. Baidu is using an AI-native video creation tool to open up a new way of playing in the video industry.
How does it feel to use Dujia in real life? What real conveniences can it bring to creators? Will it become a new AIGC hot product?
By getting started with the Dujia editing app and the Dujia creation tool PC client, Zhidongxi conducted an in-depth discussion on these issues.
1. Based on massive materials, AI generates text-to-video with one click
A video creator has these tedious tasks every day: selecting topics, finding materials, writing scripts, editing and packaging, operating and publishing... What would happen if there was an AI assistant?
Zhidongxi first experienced the new video creation method brought by Dujia from three aspects: text-to-video, AI copywriting, and hot spot creation.
1. Enter the copy to match the video and complete AI human-machine collaboration in three minutes.
Opening the Dujia Editing App, Zhidongxi first made a short news video through the Wensheng video function of the "AI Filmmaking" section.
▲Case "Musk finally admitted a low-priced new car" display
The operation is very simple - enter the "AI Filmmaking" section of Dujia, I paste the copy I wrote in advance into the text box, click "One-click Filmmaking", and I'm more than half done.
After only about 30 seconds, a video with rich material was successfully generated. The page is simple, with a video preview displayed on the upper side and editable audio and video tracks below.
Watch the video preview, the video and text match very accurately.
For individual unmatched materials or materials that can be optimized, Zhidongxi clicks "Material Recommendation" to select new videos to insert in the material library with one click.
You can also add local materials to supplement the news videos you have downloaded before, further improving the newly generated video.
Finally, after simple adjustments to the video template, pronunciation, background music, etc., a human-computer collaborative video was generated, which took less than 10 minutes.
The same experience can be replicated on the PC side of Dujia Creation Tools. Zhidongxi also conducted evaluations and obtained a consistent experience.
2. AI is all in the details, from scripts, titles to video materials, everything can be polished
In addition, AI is integrated into every aspect of creation.
For example, in the creation of the video "Appreciating the Sunset on the Chao Phraya River in Thailand", Zhidongxi used AI functions in copywriting scripts, title formulation, and video material replacement.
▲Case study "Appreciating the sunset on the Chao Phraya River in Thailand"
When Zhixi filled in a travel note written when visiting Temple Arun in Thailand into the text box, I immediately got an AI-generated video script with more vivid and colloquial language.
Click "Generate Video" to see a video that accurately captures the materials of Temple Arun in Thailand. The river, sunset, cruise ship and other images in the picture accurately match the text description.
Click Publish, I changed the title to a more topical one based on the AI recommendation - "Experience the Chao Phraya River Sunset Cruise, a Romantic Journey for 30 Baht", and then it can be synchronized to my Baijiahao account immediately, and you can also download and upload it. to other platforms.
It can be seen that AI is hidden in every detail of this software, waiting for users to explore.
3. There is no need to write any copywriting for hot spot recommendations, they are all generated using AI.
The above-mentioned films are all made with AI based on copywriting. What should I do if I don’t have a copywriting or even an idea for a topic?
Dujia can realize the entire process of "finding topics - writing scripts - generating videos" with AI. Based on this, Zhidongxi produced a short video titled "What is the Essence of Wang Yangming's Mind Theory".
▲Case "What is the Essence of Wang Yangming's Psychology" Display
When Zhidong clicks on the topic "What is the Essence of Wang Yangming's Psychology" in the popular recommendation list and clicks "Generate Copy", a script copy will be generated immediately in about 30 seconds.
I can edit on this basis, and then click "Go to create". In just a few dozen seconds, a short video with pictures and text will be generated. The video material matches accurately and can be used directly without modification.
This function is also applicable to the PC version of Dujia Creation Tool.
News requires authenticity, and source tracing is crucial to self-media creation. By clicking on the "search" symbol next to the recommended hotspot, I found that I can quickly obtain the first source of information and have a timely understanding of the facts.
For example, when I clicked "Search" next to the topic "Luo Yonghao supports Dong Yuhui's entrepreneurship" on the PC version of Dujia Creative Tools, I quickly found the first-hand source of the news and saw the original Weibo post posted by Luo Yonghao, This is useful for verifying information.
It can be seen that based on Baidu's mobile ecological capabilities, Dujia has implemented relatively mature document-to-video capabilities.
Compared with other editing tools on the market such as Cutting and Quick Shadowing, Dujia may have opened up a new generation of AI video creation methods.
When the user clicks on the hotspot, a video document will be automatically generated, and then the document will be converted into a video. A video with good quality can be generated in a few clicks. After the operation is completed, users can also check and modify details such as materials and subtitles.
The relevant person in charge of Dujia told Zhixixi that the team is currently actively exploring native video technology and will gradually implement it on Dujia products in 2024.
2. One-click synthesis of customized AI digital human, comparable to real-person interactive experience
Generative AI has also brought about the explosion of AI digital humans, which is another major advantage of Dujia.
On the Dujia Editing App, Zhidongxi saw that there are nearly ten AI digital humans with different styles to choose from. Among them are both elites in formal suits and gentle sisters next door. From the details of their facial features to their movements and expressions, they are very lifelike.
For example, Zhidongxi used the official suit AI digital human template to generate a question titled "Is it easy to drive an electric car at minus 34 degrees Celsius?" 》video. The whole process only took about 5 minutes, which is much faster than real-person recording.
▲Case "Is it easy to drive an electric car at minus 34 degrees Celsius?" "exhibit
Zhidongzhi first directly pasted the pre-written copy, and then clicked "Generate Video".
After nearly a minute of waiting, the first version of the oral video of a realistic digital human was formed.
Next, I performed an AI-assisted second creation. AI will automatically identify the keywords in the copy and provide recommendations for related materials for me to insert into the video to enrich the lens language.
For example, Dujia automatically recognized the word "Tesla" and recommended a series of Tesla-related videos for me to choose and insert. Smart Things chose one to insert.
For words that were not automatically selected by Dujia, such as "home charging pile", Zhidongzhi selected them by itself and selected the corresponding video in Dujia's material library, and played them interspersed with the digital voice broadcast.
After selecting and inserting multiple keywords into the material, a video that is mixed and edited by the anchor and video material is generated.
AI digital human creation also supports hot spot recommendations and AI copywriting.
For example, when I input a piece of copywriting with the theme of "Beijing released three important results of large-scale models" that is more official, Dujia can use AI to help me rewrite a more concise and colloquial script. After checking, I found that this passage can be used directly without modification.
As shown below, this is a generative digital spoken video.
▲Case "Beijing releases three important results of large model" display
The relevant person in charge of Dujia told Zhixixi that digital people are currently having a relatively obvious effect on reaching various industries, and users are becoming more and more accepting and aware of digital people. They are no longer just "paper people", but It is a "digital life" with more forms, personalities, and language methods.
Compared with other counterparts such as Cutout and Kuaiying, Dujia’s digital human function can already cover creators in various fields, covering a wide range of areas, such as knowledge-based creators, entertainment and hot-spot creators, beauty and medical beauty competitions Tao creators...Creators from all walks of life can use Dujia digital people to create content and let AI empower content creation.
3. Large-scale models are implemented in the mobile ecosystem, and AI native creation tools are released.
The relevant person in charge of Dujia told Zhixixi that the development process of video creation can be roughly divided into three stages:
First, in the PC era , traditional editors such as Premiere and After Effects are the main ones, and the learning cost is high. They are mainly used by professional editors, with daily active users in the millions.
The second is the App era , which is dominated by mobile products such as Kuaiying and Silhouette. The operation cost is low. Users can upload a picture and turn it into a video through a template. There is no need to learn editing skills. The global daily active users are close to 100 million.
Third, in the AI era , users only need to enter a Prompt (prompt word) to generate a video. The operating cost is further reduced, and the production effect is very creative. It is no longer limited by the user's existing pictures and video materials. Active users will reach 500 million+.
So in the AI era, what does “AI native” mean for video production?
"The key point is to make it clear that users use AI+ rather than +AI in the creative process." said a relevant person in charge of Dujia. "The AI+ we are promoting is exactly the end-to-end use of AI to improve efficiency or create creations."
Dujia AI native creators can enjoy AI+ creative support by using Dujia's product functions.
This is inseparable from the rich material support of Baidu’s mobile ecosystem. Baidu has Baijiahao's massive material library that is updated in real time, and the Visual China picture library is rich and of high quality; in terms of chasing hot spots, hot events are stored in the library within half an hour and updated in real time, covering a wide range of vertical categories, helping media people catch hot spots faster. .
At the same time, Dujia is based on the deep base support of Baidu Group's Wenxin large model, and has many years of experience in AI digital people, AI copywriting and other tracks, which makes Dujia's AI functions have better performance and a smoother user experience. , thereby achieving real cost reduction and efficiency improvement.
Conclusion: Wenxin large-scale model is accelerating its implementation, creating a new way of video creation
With one click to generate text-to-video, and one minute to synthesize a customized AI digital human, Baidu has opened up a new way of creating AI-native videos with one product. After experience, we found that the prototype of universal AI human-machine collaborative creation has been reflected in this product called "Dujia".
Behind the scenes, Baidu combined its large-scale mobile ecological materials with Wenxin large-scale models to open up the "last mile" of technology and application scenarios, thereby creating a new way of video creation. The launch of Dujia is just the beginning, and new features such as AI Wensheng native video are also being planned for iteration. We will continue to pay attention to whether Baidu can continue to take the lead in the media industry in the AI era.