I've heard of assistants doing in-between frames in animation before, but I've never heard of a computer doing in-between frames. Is it common practice nowadays to have a computer do in-between frames, or is it still mostly humans doing in-betweens? Just looking at some articles on commercial keyframe interpolation (for example, here: https://www.premiumbeat.com/blog/understanding-keyframe-interpolation-after-effects/), it doesn't look like computer generated keyframe interpolation is too powerful yet.
Haohua-Lyu
While it appears there are still no mature commercial solutions to this, some researchers and companies have looked into AI in-between generations. E.g., Adobe announced that their Sensei AI is capable of making in-betweens for photos. (https://www.engadget.com/adobe-project-in-between-preview-130017575.html) Ubisoft Montreal published a SIGGRAPH 2020 paper on robust in-betweens generation as well. (https://montreal.ubisoft.com/en/automatic-in-betweening-for-faster-animation-authoring/) And seems like some Japanese anime studios are exploring this front too!
smsunarto
More subtle tweening that has been somewhat controversial in certain animation enthusiast groups is around projecting 24fps animation to 60fps. I think it has some interesting arguments against it like the fact that computer-assisted tweening to increase framerate can lead to the animation being "less dynamic" or unable to communicate the emphasis that animator tries to put.
TianCal
I do remember watching several videos that talked about boosting a video of 24 FPS to a video of 60 FPS using AI (https://www.youtube.com/watch?v=60DX2T3zyVo). I believe they should be using same type of technologywhen ai assistants create in-between frames.
nobugnohair
Considering the principles we talked about before, I feel like it is hard for computers to adhere to these principles perfectly. For example, they might move the whole figure at the same rate, while everything is supposed to move at a different rate? To deal with this problem you would have to draw out keyframes for the transition point of every object, which then kind of diminishes the whole purpose of using AI. The follow-up details could also be missing. I wonder how they dealt with these problems.
I've heard of assistants doing in-between frames in animation before, but I've never heard of a computer doing in-between frames. Is it common practice nowadays to have a computer do in-between frames, or is it still mostly humans doing in-betweens? Just looking at some articles on commercial keyframe interpolation (for example, here: https://www.premiumbeat.com/blog/understanding-keyframe-interpolation-after-effects/), it doesn't look like computer generated keyframe interpolation is too powerful yet.
While it appears there are still no mature commercial solutions to this, some researchers and companies have looked into AI in-between generations. E.g., Adobe announced that their Sensei AI is capable of making in-betweens for photos. (https://www.engadget.com/adobe-project-in-between-preview-130017575.html) Ubisoft Montreal published a SIGGRAPH 2020 paper on robust in-betweens generation as well. (https://montreal.ubisoft.com/en/automatic-in-betweening-for-faster-animation-authoring/) And seems like some Japanese anime studios are exploring this front too!
More subtle tweening that has been somewhat controversial in certain animation enthusiast groups is around projecting 24fps animation to 60fps. I think it has some interesting arguments against it like the fact that computer-assisted tweening to increase framerate can lead to the animation being "less dynamic" or unable to communicate the emphasis that animator tries to put.
I do remember watching several videos that talked about boosting a video of 24 FPS to a video of 60 FPS using AI (https://www.youtube.com/watch?v=60DX2T3zyVo). I believe they should be using same type of technologywhen ai assistants create in-between frames.
Considering the principles we talked about before, I feel like it is hard for computers to adhere to these principles perfectly. For example, they might move the whole figure at the same rate, while everything is supposed to move at a different rate? To deal with this problem you would have to draw out keyframes for the transition point of every object, which then kind of diminishes the whole purpose of using AI. The follow-up details could also be missing. I wonder how they dealt with these problems.