Recently, China News, People's Daily, Xinhua News and other official media video number issued a work, I think everyone must have swiped to it~
As shown in the picture above, it is the video of resurrecting historical celebrities with AI production, which is very shocking. People's Daily has already harvested more than 100,000 likes, honestly, this data looks truly envious.
No matter which platform releases it, almost always the numbers won't be bad.
This installment breaks down how this video, which uses AI to resurrect historical figures, is made.
There are four steps in total:
Step 1: Historical Figures Library
We can go to Baidu, today's headlines and other news platforms above to look for pictures of historical figures, to accumulate their own image material library.
Enter the name of a historical figure, such as Confucius, click Search, and select the image.
Find a clearer picture and save it locally.
Step 2: Generate real life images with AI
The AI tool of choice here isDream.
Web Direct: https://jimeng.jianying.com/ai-tool/home
Select AI Graphics - Image Generation
Import the reference image (upload the image you just saved), the reference edge outline, the reference intensity 80-100, the selection is 95. the raw image scale selection is the same as the original image.
Input descriptor: image of an ancient middle-aged man with a calm face, eyes looking to the side, long black beard hanging down, wearing a blue turban and light-colored robes, photorealistic photography style, Ultra HD, 4K
The model defaults to Picture 2.0, with the fineness cranked up to the max.
Choose an image that works better and save it locally.
Step 3: Generate Video
The tool of choice is vidu, webpage: https://www.vidu.cn/
① Select Tusheng video, vidu2.0 model, upload character portraits and real life pictures.
Descriptors can be entered: The first picture slowly turns into the second picture, and the figure in picture 2 is stationary
The generated video is first saved locally.
② Next, just upload a picture of the facsimile.
Cue word: Subject's eyes look into the camera, smile, and nod slightly.
The generated video is saved locally for use.
Repeat the above and probably do 4-5 characters a video and you're done.
Step Four:Clippingedit a film
Open Shear Image, import the videos you just made in order, and import several groups of videos if you have several characters. That is, import the ① and ② just now.
Just add text, add background music, and set a cover to export the video.
Lastly, thanks for watching the whole tutorial~
