Hi rendalunit,
you're right!!. It worked also in my script:
REN
I never tried "lanzcos" in up or downsizer. Does it look the image sharper?.
And about your sample code, I'm with Jellygoose: I substituted temporalsoften and unfilter with mipsmooth (I posted here the code, I'm using it instead the adaptive part of MA optimal script). And mipsmooth is the last filter (just msharp after once in a test, when it got too blurry, but I'm lowering MipSmooth parameter values now, to avoid having to sharpen: don't like the result, and bigger sized files). Even when I test vaguedenoiser with mipsmooth, I leave mipsmooth the last cleaning-denoising filter.
If I like the result, I'll post here "my" (better "our", because everybody in here is helping) adaptive code with vaguedenoiser and mipsmooth (and maybe I'll post it anyway, because somebody may find interesting working on it, even if I'm not able to fine tunning it). Has anyone an idea on how to adjust MipSmooth to adapt to motion (more temporal on low motion, more spatial in high motion, and so...)