Blackmagic Forum

Am I right when think Fusion does NOT work in resolution independent manner like "smart" objects in Photoshop or bitmap objects in vector soft?

I mean if you downscale something and then upscale back you will get a blurry mess. Or if you have to do second crop and downscale it results in a slightly worse picture comparing with Photoshop?

So I always have to keep in mind where the scale node is in the node flow?

Just tested it in Blender out of curiosity, downscale and then back upscale an image and looks like it does it more or less "smart" way like Photoshop


ps. oh, looks like it depends on what tool does the scale. Transform is "smart" Resize and scale are not

weird..
in fu 6.4 transform, scale nad merges do the "smart" scale, resize indeed do not, adding just a tiny little bit of softness (while doing x10 operations -> scale/resize/transform) .
Also do remember that any filter/effect between those nide will destroy the scale info (ie color corrector, or blur). So basically do the CC/fx before or after the scaling/repositioning.

EDIT: checked this in Fu7.5 - still works, whether it is a text node, or a photo.
Also check filter options in those nodes - Lanczos is pretty much the best when scaling resizing (IMO).. might not be for animation though..

hope this helps.
happy fusioning

Thanks for the tip. Looks like the resize should be only in the end of the chain

You should always try to put several transform nodes one after another if it is possible. That way there would be only one real transform (and only one image incarnation of image degradation). The same is applicable to colour corrections. There are no "smart" transforms in the means of Photoshop in Fusion though. Unless you work with text, masks or 3D models which are vector based (as opposite to pixel based images). But you always can load higher resolution images into the loader and Fusion will recognize and treat them properly (masks and transforms are percentage based as opposite to image resolution based). I hope I was clear enough. Good luck

There are certain exceptions, namely tools which are specifically made to work with pixels.
Resize and Crop come to my mind.

Instead of Resize use Scale. As for crop there is a Fuse called CropRectangle which works in relative space.

Every "Creator" (think background etc.) has a resolution tab. Be aware of this. You can use the "BG from Image" Fuse if you want to create a solid color with the same resolution as an upstream tool.

You can also use simple expressions to hook up all creators with each other, or use scripts to change the resolution afterwards. It is more convenient though to set your initial resolution in the comp preferences.

One last note: Specific Subsampling Filters only takes place in HiQ mode for most tools in Fusion. So if you do not see changes in the Subsampling check if you enabled HiQ, or disable the "Only use Filter in HiQ" if available.

resize, scale, corner positioner and letterbox don't concatenate with each other or with transform tools. I might be wrong but I think the tracker doesn't either.

last time i checked (i think yesterday or so) scale and transform do concatenate..(in lanczos mode scale x10, xf size 0.1)
do not know about the tracker, but i suppose it should (especially when in "matchmove: mode, since it would operate as merge; and merges work with xf...)

What you are talking about is non destructive pixel transformation, aka smart object in PS. Fusion is non-destructive by nature due to being a node based unlike Photoshop where you have to make copies of layers in order to manipulate pixels in case you want to preserve the original. Even in CC 2014 you still need a script revert smart object back into a layer. Smart object is the most powerful feature that Adobe refuses to evolve and abuse, that's why Fusion is a lot better solution when it comes to image manipulation and compositing than Photoshop. Due to Fusion ubber masking ability I am much quicker to make and modify matte paintings than I am in PS, you could argue its personal preference or workflow but when it comes to actual versioning (like night, dawn, day) Fusion kills PS in that matter. You could spend almost an hour to arrange proper layer compositions while in Fusion its only a matter of few clicks.

Anyhow I think you are looking for a non destructive scaling which would be your Transformation node. Resize mode should always be at the end of the workflow in case you need to export to proxy or smaller resolutions. True you can use it to bump up the image size in the loader to match your output resolution of other content but that would be a wrong setup from the get go in my opinion. Unless its a little supportive piece to help the comp. No matter if you use Transformation or Resize/Scale make sure you go over different filters. In my opinion Lanczos works best universally but Mitchell or Bessel might do the job better sometimes.

In my mind resolution independence is something more of an app being able to save all the previous adjustments correctly pixel for pixel when you change the actual resolution of the manipulated image.