Dear Emma,
AFAIK compute the IR spectra through dipole (Berry phase), Wannier or Voronoi integration must result in almost the same spectra, I remember Dr. Brehm shows the unique difference between them is the relative intensity between peaks.and Wannier could be problematic sometimes when there is high symmetry. So there is no problem. BTW, you can compute the properties of your trajectory using the REFTRAJ option in the MD.
About the IR spectrum on TRAVIS, it can be computed with Voronoi integrated electron density (which currently can be done directly in CP2K on the fly), with fixed charges or fluctuating charges calculated along the AIMD. I used the last one with self-consistent Hirschfeld charges and it worked very well.
But the point is, with just your trajectory you can compute the power spectrum which can show you some "missing" signals in IR (maybe), and the NMA without problem. The other spectral properties need electron densities and more.
In the program, the NMA is called NC (normal coordinate analysis) . When the "List of functions" is displayed, in the Spectroscopic Functions section there is "nc". This is what you need. Then you need give some information and feed with the conrformer's structures (previously optimized at the same level).
I suggest computing first the power spectra with TRAVIS and "optimize/play" with the variables and become familiar with them (the defaults are good, but sometimes a little change can upgrade the spectrum), because for the "nc" the same procedure is done + the reference structures.
Finally you will have something like this:
Mode Integral (K) Center (cm^-1)
----------------------------------------------
1 -0.112446 2735.82
2 -0.0757895 1377.51
3 -0.0508332 529.47
4 -0.00656595 615.81
5 0.0120252 542.87
6 0.177628 1415.45
----------------------------------------------
7 327.5 528.37
8 335.364 535.27
9 338.557 561.62
...
23 319.248 2958.93
24 330.046 3163.23
----------------------------------------------
Which list the modes frequencies and their thermalization... The mode temperature must be similar to the AIMD mean temperature (the first six must be near zero)... and a molden file is printed with the normal modes for each reference structure where you can see them. The normal mode frequencies must be similar to the peaks in the power spectrum.
In my experience when a time step over 1fs is used the nc starts to fail... also if you have the velocities "trajectory" you can use it (with -vel option) to obtain a better velocity autocorrelation function when the time step is equal or bigger than 0.5fs.
If you have doubts or need help do not hesitate to write to me, I am not an expert but I can help or learn on the way :P.
Regards - Lucas