@vravn Well, considering that motion vector calculation is a fixed load per source frame, independent of how many intermediate frames you insert (each one being cheap to compute), it would probably be possible to make it "realtime" if the output video ran at a normal framerate with N frames inserted between each original. My current implementation is way slower in any case, taking almost a second per processed frame. Of course I've spent very little effort on optimization though.
Impressive work DrPetter! Normally i walways thought slow motion was only attainable at a decent quality when using those scientific high speed cameras that work at incredible frame rates.
Fast motion (like in normal 30 fps video) doesn't currently track very well with this method/implementation, so I'd have to do some more work before that's practical. I think it would be reasonably efficient to simply do it in multiple steps though, starting with a downscaled version of the image to track larger movement and then progressively refine it using higher resolutions.
I came up with it all on my own, but I suspect that the general concept has been known and used a lot before. Most simple and effective things have. I'll try to get an illustration up on my homepage at some point. You can keep an eye out for it.
Will you release the soft ? I need for a short movie i record :) (Ark my english is too bad !)
Coldragon78 7 months ago
That looked so sore -.-
madlazza 8 months ago
God damn that music is awesome.
RazputinOleander 1 year ago
What program did you use for that?
JARMOPOWER 1 year ago
@JARMOPOWER its a camera test you idiot. read the description, is a HIGH SPEED CAMERA.
madlazza 8 months ago
@madlazza Oh fucking shut up if you can't read the description -.-' he SAID HE USED A PROGRAM TO INCREASE THE FPS YOU IDIOT.
JARMOPOWER 8 months ago
@JARMOPOWER no he fucking didnt
madlazza 8 months ago
@madlazza "with additional time distortion added using an application I wrote over the last two days. "
Everyone wants to be a cool guy on youtube but it seems that you've failed.
JARMOPOWER 8 months ago
@JARMOPOWER Why would i try to be the 'cool guy' on here. Im the cool guy in reality xD
madlazza 8 months ago
@madlazza Let's just agree that both of us are cool. :) /conversation
JARMOPOWER 8 months ago
@JARMOPOWER i liked that comment because its true :L
madlazza 7 months ago
I need it URGENTLY! Poor animators best reference buddy!!! Cheers Petter ,)
JoeQwesto 1 year ago
Impressive!
revengeofmonty 1 year ago
drpetter is a genius, that's about all I can say at this point.
alexkemmler 1 year ago
This is pretty sweet. Fast enough interpolation to be used in realtime?
vravn 1 year ago
@vravn Well, considering that motion vector calculation is a fixed load per source frame, independent of how many intermediate frames you insert (each one being cheap to compute), it would probably be possible to make it "realtime" if the output video ran at a normal framerate with N frames inserted between each original. My current implementation is way slower in any case, taking almost a second per processed frame. Of course I've spent very little effort on optimization though.
drpetter 1 year ago
Impressive work DrPetter! Normally i walways thought slow motion was only attainable at a decent quality when using those scientific high speed cameras that work at incredible frame rates.
nanakimurasaki 1 year ago
nice
pedobear123 1 year ago
Cool! Hehe, almost got the trick too ;)
ScaroDj 1 year ago
Fast motion (like in normal 30 fps video) doesn't currently track very well with this method/implementation, so I'd have to do some more work before that's practical. I think it would be reasonably efficient to simply do it in multiple steps though, starting with a downscaled version of the image to track larger movement and then progressively refine it using higher resolutions.
drpetter 1 year ago
Could you make one based on a normal-speed video and perhaps with before-and-after? It looks magical, and I'd love to see more of it!
Magnulus 1 year ago
I came up with it all on my own, but I suspect that the general concept has been known and used a lot before. Most simple and effective things have. I'll try to get an illustration up on my homepage at some point. You can keep an eye out for it.
drpetter 1 year ago
Cool, have you got any more info on the analytic method you used? Is it entirely your own or based partly on a paper or something?
Nice vid : )
tobuslieven 1 year ago
Cool effect ! :)
rockjas 1 year ago