Russia
Joined: Aug 13, 2022
Post Count: 636
Status:
Offline
connecting timelapse SH3D and video AI
I have a new interesting task. In our yacht club, we need to decorate an ugly concrete pole for electrical wiring. first, I created a model of cladding for a concrete pillar from shaped panels. I placed the office containers on a wooden floor, adjusted the lighting using hemispheres of different colors and power. Next, I chose an image of two people under an evening street lamp.
I cleaned the background of this image and saved it as a PNG texture. I placed it on the front of the box and made all the other sides of the box invisible. The result is these renderings, which I have numbered for you to make it more convenient. 1 and 2 with these people, the remaining 3.4 without people and 5 with a dog.
these renderers will be the basis for the video. In the next answer, I will tell you in detail how to do everything.
Russia
Joined: Aug 13, 2022
Post Count: 636
Status:
Offline
Re: connecting timelapse SH3D and video AI
now how to make a video using AI. This is the video transition mode from frame 1 to frame 2. The text task is "the girl tilts her head and looks at the guy". The next video clip is the transition from frame 2 to frame 3. The text task is "A girl and a guy get up and leave". I received two videos of 5 seconds each (the process of creating such videos takes about 3 minutes in AI, the renders themselves take 3 hours in good quality). Next, for the new video, I connected SH3D in the latest version of the settings, where you can set the curved arc between the camera positions, set up the camera passage from the position of frame 3 to frame 4. Frame 4 is the camera position, not the render. I will receive this frame only after completing the process of creating a video in SH3D from position 3 to position 4 (I will take the last frame from the video as image 4). I set the first position to 4 a.m. (before dawn). I set the last position of the camera with the camera height rise to 22.00, after sunset. Between these positions, the sun illuminates the scene with successive rays from different angles, and the shadows move dynamically in timelapse mode. From darkness to darkness every other day. This process in good 4Q quality took (oh horror) 76 hours of continuous computer operation. But with the help of AI, you can only do something like this, I tried, SH3D does everything efficiently, but for a very long time.. I had a lot of free time, so I decided to do it. I made the initial mistake. I set the frame size as a square for the first renders. But in the SH video settings, there are only 4:3 and 16 sizes as 3 and 4.:9. Therefore, in the final video there will be a strange transition from a square to a rectangle in the frame. This should be taken into account when working with SH videos. Start immediately in 4:3 or 16:9 format. After a delightful timelapse of SH, I took the last frame of the video and assigned it as the first frame of the next video. I quickly made this fragment again with the help of AI. Frame 5 is the view with the dog, frame 4 is the top view from the SH video without the dog. The text task is "a dog comes out of the dark with its tongue hanging out and sits down, looking in different directions." That's all. It turned out to be an interesting video without unnecessary fantasy distortions. The SH3D project only. all that remains is to stretch the video fragments a little to the music. video at the yacht club
