Hi.
Great work!
I am an amateur in making funscript. I used to use OpenFunscript. I've been using FunGen for a while, and I found the AI track generating function might not be that good than expected. Here are some comments based on my experience making funscript with OpenFunscript.
OpenFunscripter uses MTFG to generate tracks (surge, roll and other six axises). Although it can be obsolete, OpenFunscripter utilzes MTFG in a much smarter way, involving human recognition which enables precise locating of critical areas, while current AI track generating is disturbed by multiple objects detected in the same frame. The options in current AI track generating is barren too.
Although the MTFG is often interrupted when target moves out of the view, carefully choosing tracking target (e.g. in certain case the navels of both characters are always in the frame) can lead to fluent tracking and yield more robust output than current current AI track generating method. However, this trick can't be extend to most 3D video or video contain complex view changing. It might be promising to fix this flaw with AI.
Best wishes.
Hi.
Great work!
I am an amateur in making funscript. I used to use OpenFunscript. I've been using FunGen for a while, and I found the AI track generating function might not be that good than expected. Here are some comments based on my experience making funscript with OpenFunscript.
OpenFunscripter uses MTFG to generate tracks (surge, roll and other six axises). Although it can be obsolete, OpenFunscripter utilzes MTFG in a much smarter way, involving human recognition which enables precise locating of critical areas, while current AI track generating is disturbed by multiple objects detected in the same frame. The options in current AI track generating is barren too.
Although the MTFG is often interrupted when target moves out of the view, carefully choosing tracking target (e.g. in certain case the navels of both characters are always in the frame) can lead to fluent tracking and yield more robust output than current current AI track generating method. However, this trick can't be extend to most 3D video or video contain complex view changing. It might be promising to fix this flaw with AI.
Best wishes.