Hey everyone!
I wanted to share a little behind-the-scenes look at what Iโve been experimenting with for Act 2. Lately, Iโve been working on making longer, smoother animations. My current process goes something like this:
Create a short 4โ6 second animation
Save the last frame of that animation
Animate that last frame as a new sequence
Use FFMPEG to piece them together into one continuous clip
This works pretty well, but it also creates some challenges:
Color shifts over time can happen between stitched parts
The last frame doesnโt always have the same detail as the first (eyes closed, head turned, etc.)
When the character moves back into frame, details can sometimes change unexpectedly
Hereโs an example:
First image (eyes open) โ initial animation starting point
Second image (eyes closed) โ the last frame of the 6-second video that I then try to animate
In the second image, certain details are missing or obscured (like eye color, facial features, expression, even the look of the man's penis). Since thatโs what gets input into the model for the next sequence, it can cause the generated animation to drift away from the original look.
Another thing to consider is that this approach adds significant development time. Each sequence requires more trial and error before I can get a satisfying result. That means I can either create more shorter (4โ6 second) videos, or focus on fewer but longer videos that take much more time to produce.
Iโd love to hear your thoughts:
Do you prefer more frequent, shorter loops?
Or fewer, longer animations even if it means a slower pace of development?
This post may be a bit preemptive, but youโll get to see the results of my efforts in the upcoming build this week!
โ MorrakGames
Mike
2025-09-07 08:14:42 +0000 UTCBRobert
2025-09-06 22:29:30 +0000 UTC