Face changing AI computing power graphics card
Take fakeapp as an example
1. Due to the high requirement of AI face changing software for GPU, it is required that the display memory of the graphics card should reach 4GB and the memory should be at least 8GB
PR can change face, AE can also
patiently put the face close to the skin color into the tracking frame by frame, you need to fine tune the position all the time, and the size should change with the change of the face
if it's a cross between two faces, it's cute to use AE reflex plug-in
AI can change face? You can't believe that people will give you videos later
recently, a reddit user put gal gadot's face on an alt video, and the effect is quite good. What's more, he uses the popular machine learning technology
The key of
motion capture technology is to recognize people's expression, especially those key points on the face. And deep learning technology can also meet this requirement. Some research teams even demonstrated fake videos of former US President George W. Bush, Obama, current president trump, and Russian President Vladimir Putin
compared with those special film and television instry technologies, the results of deep learning technology are much coarser and easier to identify, but it is also much cheaper. The software technologies and platforms of big companies such as Google, Microsoft and Amazon are open. Deep learning technology is developing rapidly, and many papers are shared among researchers. In addition, home level graphics cards can handle deep learning algorithms, and do not have to use cloud computing and supercomputer technology< br />
The essence of AI face changing is to separate facial value from facial expression; The expression is the image of the facial features in different emotions, more refers to the movement of the facial features
However, the appearance of AI face changing technology is telling us that facial value and expression can be separated The final merged effect of
is the expression in the original video; The facial features become the facial features in the photos, and the result of face changing is very natural in the whole video clip
in the video generated by this process, facial features come from pictures and facial expressions come from videos. Facial features and facial expressions are separated, that is, facial values and facial expressions are separated
In other words, the essence of AI face changing may be the separation of facial value and expressionWhat's the use of separating facial values from facial expressions
revolutionary innovation of film and television proction:
in film and television proction, there is a "separation" that rewrites the whole film and television instry, that is, the separation of image and sound
at first, the film and television instry was silent film and television. Later, some people creatively played the sound and image simultaneously, and the sound film appeared
and the separation technology of sound and image has been used until now. This separation ensures that the sound can be proced independently, creating a sound effect with stronger sense of three-dimensional and scene
and the separation of facial value and expression will have a stronger impact than the separation of image and sound
(1) greatly increase the facial expression of film and television works
for example, in the past, it was hard to imagine that people could come out of "painting", but now through AI face changing, as long as you get the facial features in the painting and put them into any video, you can make a video of people moving in the painting
we need to know that the face in the painting is extremely free. We can create any face we want with our free will, and then change the face through AI to create any vivid character we want
Methods of using AI face changing software:
1. Open Zao fusion, take photos with mobile phone or select the replaced photos from album