Hengyu originated from the Cave Fei Temple
Qubits | Official account QbitAI
Pay attention, this video of the robot fighting humans has completely gone viral on the Internet.
That is called a fist to the flesh, and the degree of stimulation caused the onlookers to shout goose girls.
The good news is that this robot is not real.
But another piece of news that I don't know if it's good or bad is that such special effects are synthesized by AI with one click.
That is to say, from the human real shot on the left to the special effects picture on the right, all humans need to do is drag the model into the picture with the mouse.
This AI artifact is called Wonder Studio AI.
When the promotional video was first released, it caused a lot of wo cao.
Now that the internal test quota has begun to be distributed, netizens can't stop playing and directly brush the screen on social media.
For example, a boy band dance that is more dynamic and smooth than Boston Dynamics.
Or freestyle in your own room.
↑↑↑The author of this work also revealed that this is a visual effects shot that originally took three days to complete, and now it only takes 1 minute.
No wonder netizens can't sit still.
No need for motion capture equipment, one-click CG animation generation
Regarding Wonder Studio, the official version is, "an AI tool that automatically animates, lightens, and synthesizes CG characters into real-world scenes."
According to the official website, for users, in the CG animation production stage, you no longer need complex 3D software, nor do you need expensive motion capture hardware.
All you need is a camera to shoot, shoot a video and throw it into Wonder Studio to process, which is enough to fake the real thing.
The specific steps are as follows:
First, a video is shot with a camera, in which the live actor can make any movement at will;
Import videos in MP4 or MOV formats into Wonder Studio, select the video segments to be processed, and drag them into the middle area to wait for processing;
Click the next button in the menu bar, and the AI starts to identify and analyze the people in the video;
Select the CG model required in the option bar on the right and drag it to the character area to be replaced;
At this time, a small box will pop up, and you can adjust HDR lighting, image quality clarity, video format, etc. as needed.
Finally, select the solution generation to complete the replacement of the CG model to the live actor in the video paragraph.
For professionals, in addition to the final product, Wonder Studio can export different tracks of new videos, including motion capture, camera tracking, empty mirrors, and more.
In the face of details that need to be modified, you can also directly start fine-tuning.
The traditional CG production process requires removing the original characters, importing the CG elements that need to be replaced into the engine or software such as Blender, compositing according to the camera solution results, matching lights, rendering... Finally, the new CG model is put into the video scene.
Comparatively speaking, after using Wonder Studio simplifies many processes, it is no wonder that many people comment that if this workflow runs through, it is very conducive to small-cost film and television production.
Wonder Studio isn't perfect, though.
It is mainly reflected in the fact that after the CG model covers the original characters, the CG characters and background articulation are not excellent.
For example, there will be a situation where the beauty and face slimming effect is so large that the background around the face is distorted:
Moreover, the edge pixels of the CG model will be stretched, and the original character's clothing cannot be 100% eliminated and covered:
In the explosive video shown at the beginning, there are actually many gang shots.
Therefore, although the effect is powerful and simple to operate, for now, Wonder Studio cannot be used alone in actual projects, otherwise it will be amazing at a glance and then look at the stuffing.
But the potential it has shown has been affirmed by internal test netizens.
According to feedback from beta users, in several of his tests, Wonder Studio was able to automate 80% of the VFX work.
It still has some difficult areas, but the value proposition it offers is that it takes very little effort to make something pretty good.
And this is just the beginning! It's going to get better!
"Ready Player One" male protagonist entrepreneurial project
It's not just the products that are interesting, the company behind Wonder Studio also has a story.
It is produced by an AI startup called Wonder Dynamics, which aims to improve the efficiency and quality of filmmaking through AI.
To date, the company has raised a total of $11.5 million in seed and Series A.
Wonder Studio is an AI-based online video effects processing platform launched after 3 years of preparation.
The company has said that Wonder Studio will continue to simplify motion capture and rendering and continue to improve production efficiency. And in the company's vision, creators will be able to freely upload their own 3D models to the system in the future.
Looking at the founding team of the company, many people are familiar faces.
For example, the world-famous director Spielberg is a consultant to this company.
And one of the founders of the company is "Ready Player One" actor Tyre Sheridan.
(I don't know if it's good to say that the male protagonist "does a line of love", or that he is "actively committed to using AI to grab his own job",
）
It is also reported that in a film being shot by the American director group Russo Brothers, Wonder Studio has been involved in part of the production process. The film stars Stranger Things star Millie Bobby Brown and Star-Lord Chris Pratt in Guardians of the Galaxy, and the film will be available on Netflix.
Before looking forward to the movie, it is more noteworthy that in recent times, the power of AI has been crazy in the visual field.
This has to mention the Meta killer that has recently been frantically brushed in the technology circle, Segment Anything Model (SAM).
The model is amazed by its ability to identify and separate specific objects in images and videos. With SAM, users can select any object that needs to be edited by clicking or entering text.
Meta wrote in introducing SAM:
SAM has learned general concepts about objects, and it can also generate msaks for any object in any image, or video, including types not included in the training set.
SAM is versatile enough to be used out-of-the-box without additional training in new image areas.
This ability to migrate zero-samples is one of the important reasons why GPT-4's capabilities are amazing. Therefore, many media at home and abroad called the release of SAM a ChatGPT moment in the CV field.
Currently, SAM is open source and further provides full details of its 1 billion mask dataset,
Meta's official statement is that as technology develops, SAM can become a powerful component of more general systems such as AR and VR.
Thinking of the Wonder Studio introduced today, after accessing the SAM capability, the segmentation of the original human image, will there no longer be a situation where the skirt corner is missed?
Closed beta application address:
https://docs.google.com/forms/d/e/1FAIpQLSdLY3_MQuMEqBy-MZCyXHoPe2YK5BUiSQecjCt0eLUQCux4qA/viewform
Reference Links:
[1]https://twitter.com/shota7180/status/1644815889576128512
[2]https://twitter.com/wonderdynamics/status/1644376317595615233?s=46&t=HBob6gxh8cOfZTIbieKeSA
— End —
Qubits QbitAI · Headline number signed
Follow us and be the first to know the latest scientific and technological trends