DAINAPP
patreon
A few examples for the new model.
🕑 Added 2021-09-24 00:20:34 +0000 UTCOk, some stuff first:
- I found a bug in the model, so I'm training again. On top of that I'm trying some extra stuff that I been studying, so tomorrow I may have a better model? Hard to say with AI.
- This examples are using only Step 1 of the model. There is some advantages and dis some good and bad on that.
Good:
- Using only Step 1 you can interpolate 2X 3X 4X 5X 6X 7X 8X [..] 99X with almost no extra time. The extra time is pretty much writing the extra frames on the HD/SSD
- Its faster than using with Step 2.
- Can interpolate 3X 5X 7X etc
- Some times is work better for animations than using Step 2 as well.
Bad:
- Interpolation more stiff, remove some of the flow on new frames.
- It can generate more distortions and bad looking frames.
- Impossible frames, like a arm changing the location by a lot look worse.
Examples in the following posts. (Stupid Patreon don't let me put all in here)
Comments
GRisk
A masking system is not that hard. A System like that with give good results it should be extreme hard. But nothing is impossible,
MindChamber
Agreed, I was more inferring to the ability to lasso parts of an image.. say a leg or fist at different frames and tweak the ride app to under stand the transition better by cutting the image of a frame and moving it in a position where the artifacts are happening
GRisk
It look like a sowtware that work with vectors. The best solution would be to create a AI that turn images into vectors. This would be no easy task.
MindChamber
https://www.youtube.com/watch?v=bvxYGOmnvpM this program is obviously more involved, but its an animation software that allows tweaks for in between.. I wonder if there was a way to tweak or clean AI tweens in a similar way
