Thalin DeShaker experiment - why didn't this work?

[brassplyer]

Using the Thalin DeShaker in VirtualDub I went through the first step - video analysis pass which writes to a log file. When the analysis pass finished instead of then going to step 2 and saving I exited out, then reloaded the file and *then* did the second pass. The "DeShaking" message happened which I would assume means the analysis data is applied to the file but the end result after subsequently saving the file was a mess.

Why wouldn't this work? Why would the log file care if the file got closed and re-opened later if the analysis is based on the exact same file?

[poisondeathray]

Possibly:

1) frames mismatch e.g. maybe script wasn't frame accurate (directshowsource) , or sometimes temporal filters, or MT can sometimes cause it

To test this, use a lossless intermediate for the 1st pass (don't stack a dozen filters on the passes, especially temporal ones), and try a frame accurate source filter

2) wrong settings for deshaker, or wrong type of stabilization (not all types of scenes are amenable to deshaker type stabilization), or certain scenes can fool the deshaker analysis (eg. like big object crosses in the scene, can skew the data)


BTW , you don't have to "exit out" or reload the file when using deshaker, you just go to the 2nd pass directly (or was that your "experiment") ? I've done this before too, and it works fine when closing vdub and reloading it later on

[brassplyer]

BTW , you don't have to "exit out" or reload the file when using deshaker, you just go to the 2nd pass directly (or was that your "experiment") ?

Right, I just wondered what would happen if I tried it.

[budwzr]

What is "MT"?