I don't think its a wise idea to mess with the trajectory.
What you should do is when loading your trajectory in VMD, you can skip any number of frames. So for example in the selection box (see attached) you can have
first: 0
last: -1
stride: 1000 and that will load every 1000th frame from your trajectory.
You can of course change the stride value to suit your need.
Use trjconv to extract say 2 frames for every 1 ns. That is you need to extract data every 500 ps. The number of frames you skip will depend on how frequently you write the .xtc file as specified in the mdp file.
After that use trjconv accordingly to say make a pdb file of the simulation. Then play it in VMD. Use movie maker / screen cast software afterwards.
I don't think its a wise idea to mess with the trajectory.
What you should do is when loading your trajectory in VMD, you can skip any number of frames. So for example in the selection box (see attached) you can have
first: 0
last: -1
stride: 1000 and that will load every 1000th frame from your trajectory.
You can of course change the stride value to suit your need.
I use pymol for HQ pictures. Simple for loop in the bash that loops over the input pdb and calls a script for the rendering. Thats more stable as compared to load the whole trajectory in my experience.