New composite filter nodes

General Forums Blender and CG Discussions
#1

Hi everyone.
For our design project, 2 friends and I are planning to add some composite nodes to Blender and we are currently evaluating some nice filters from Gimp. We would like to know which one would be more useful and which you would want to see in Blender. Here’s the list of the filters that we are considering :-

Thank you all for your comments. We’ll keep you posted with more updates.

#2

Newsprint would be nice.

Cubism Wind and pixelate would be ok if their settings were fully animatable.

I would like a Radial Speed blur tool. for example making it look like you are going fowards rather fast if rendering from the front of a race car or spaceship (without using Vector-blur and putting in Keyframes).

I find many of those filters would not be useful in many situations at all. perhaps they can all be put into a single “Filter” node which has a drop down menu for the different types of image filters that can be applied, rather than having 10 different nodes for 2D image filters.

Sounds like a good project though.

Doug

#3

Most of those look a bit gimmicky to me, and not all that useful for practical work.

I’d personally say that Deinterlace would be useful, as well as Pixelise and one of the two noise filters.

If I could give a suggestion, I’d encourage finding out what’s needed in Blender, and analysing workflow, comp setups, etc to see what’s actually useful in a practical sense, and what can help make things easier for users to get done the sort of tasks that Blender’s compositor is actually used for. You’d even be much better off looking at other similar compositing apps like Shake or Fusion (even reading tutorials for those apps!), and seeing what sorts of tools they have that are useful for digital compositing. Just copying novelty filters from gimp (which isn’t really the same class of app as Blender’s compositor) doesn’t seem that good a use of resources to me…

#4

you might be a lot better off leaving the gamut of artistic effect filters–which are fairly static and rarely used INSIDE photoshop and the GIMP, let alone outside–and including more technical functionality filters, adjustment filters, camera effects and mini image-operations that can be chained together.

#5

Artistic filters may be very useful if you allow for the full range of artistic effects from watercolor to oil to pastels to crayon, sliders to control influence, animatable, and ability to mix and match them in any combination with any other nodes. Though you may want to keep it to one very powerful node if it’s possible to minimize node clutter in the menus.

If you could do these for the material nodes then that would be even more useful.

#6

Speak for yourselves! I would very much indorse the inclusion of any more artistic tools. I think that adding painterly effects within the composition stage is a great benefit for animators. Sure you could do this stuff externally but its a real pain for image sequences.

And yes it would be cool for nodes too. But I’d settle for comp’ing

#7

I have to agree with the above, artsy filters don’t seem to be very useful to me, especially not in the compositor.

This is just my opinion, but things like Cubism, Cartoon, Oilify, Photocopy and Newsprint would be much better created using the already existing nodes. If a node that’s required for such a filter doesn’t exist yet, I think it would be a better idea to just create those nodes that would be required to do the full artsy filter.

It would add more flexibility than hardcoding every artsy filter out there, even more things could be made with it, and the menus would not become crowded with filters that aren’t used very often.

Things like fractal trace would be better in the material nodes I guess. Make it an input node, like textures and colors. Or even better yet, convert the whole noise, etc.-system from the buttons-panel to a new input node, so there’s less reason to switch to the button space when using nodes to create layered materials :slight_smile:

</brainfart>

#8

I’m with BlackBoe. Given an exclusive choice between more variable input nodes and complex fx nodes, I go with the former.

#9

(edit: I thought I was debating with CubOfJudahsLion but reading it again I think we’re agreeing :stuck_out_tongue: Anyway, my rant below still stands, IMO).

The question is: How valuable are the effects that those inputs produce? Is this an efficient use of effort to help Blender users get their work done? Most of those cheesy ‘artistic’ filters (for want of a better word, since their outputs are paint-by-numbers amateurish at best and don’t really have much to do with artistic expression) just don’t get used much.

It’s not really just a case of “add everything! more more more!” since adding things to the source has a maintenance cost - fixing bugs and if architectural improvements are made, it’s more legacy code to update, and also in usability - if the menus are full of these sorts of things, it’s harder to find the useful things that get used all the time.

How many people are really going to be using a newsprint node? Are we all going to start making gaudy 1980s music videos? And ditto for the others. We just don’t need specific nodes that one in a thousand people will find useful, one day a year.

That’s why I suggest actually doing some analysis to see what’s needed. Even something as simple as taking the existing texture nodes, and improving them so you could have (for example) a separate node for each texture, with node inputs into all the texture properties so you could control them from the node editor rather than the awkward texture panels would be a huge boost in functionality and usability. These are the sorts of bread-and-butter things that need to be done to actually help people’s workflow, not just something that you’ll play with once, and never find practical usage for afterwards.

As a side-note, if we had a plugin system for nodes, this wouldn’t be an issue. You could make node plugins and use them if you wanted, without having to maintain them within blender. Some initial work was done on enabling this, but it’s not ready. Perhaps that would be a good thing to work on, too.

#10

By all means, a study would work. But don’t forget to consider the opinions posted here as part of it, as it might reveal a trend.

#11

Broken: Yes, I think we’re agreeing. Solid dissertation, though. :stuck_out_tongue:

#12

I agree that these “gimicky” effects are not really necessary, and most of them can already be acheived with existing nodes.

I would like to see more work put into managing node groups, so they can easily be packaged up and re-used in other scenes, with only the relevant sliders appearing on them, labelled for what they do. That way node setups can be made more user friendly for re-use and sharing with colleagues.
I know this can basically be done now, but the interface needs tweaking IMO.

#13

yes, but you know, there are always people out there that want to add hideous superimpose templates to photos too…
it really would be great to have nodes as plugin system… this “is this truly important for blender”-fight with devs people have to go through everytime is a pity.

there has already been couple of nodes that i would’ve love to have, but never got implemented “because you can already do this in blender, in a much much harder way”.

.b

#14

I agree that these artistic filter are rather for static or 2d. For this I would preferably use specialised 2d applications such as GIMP which always be better. For 3d stills I would always pick specialised 2d application for post production fork flow.

For animations these filers will hardly be used regularly. That is why it would be better to analyse which specific filters (used in other compositing applications) are really needed by users and for 3d animations (where 2d application is not really best tool for compositing work flow).

#15

Guys I like your readiness to add to blenders capabilities. Always good to have your sort of generosity around here. Thank you in advance.

Are you guys the same team that were spoken about on the Sunday Meeting

keirs students are working on composite nodes - one such is a refraction node - http://blender.astrometry.net/wiki/Refraction

they proposed implementing the following:
http://blender.astrometry.net/wiki/ComicBook http://blender.astrometry.net/wiki/wind http://blender.astrometry.net/wiki/InpaintingNode http://blender.astrometry.net/wiki/MosaicNode
which seemed to be supported as worthwhile goals.

If not then maybe you could find out what help that team needs and how you can help them, and if you guys are the above mentioned students, then it appears you may have your foot in the door.

I do like the idea about making the node system plugin system as there will be endless node possibilities that could be added if needed, and if not, then it doesn’t have to be there if you don’t want it to be.

Check out what needs to be done to get that plugin system working and you will definitely have our support.

As stated above thank you for your hard work and generosity, and good luck to you.

#16

That’s the way, aha, aha… :yes:

#17

There is a plug that called fblur in gimp. It a focus blur plug that seems to be more correct focus blur then that one in blender that exists. It’s painfully slow but makes much better depth of field.

#18

… Puts hand up as one in a thousand…

I personally don’t see the harm in a single node called “Artistic” that loads many of these types of filters.

How do artistic nodes compare to non-photorealistic renderers by any chance? Non-photorealistic renderings are very much desired in many industries as a tool to communicate the conceptuality of something.

#19

… Puts hand up as one in a thousand…

then more hands will follow…
I like the idea.
What I would really like would be the ability to load Gimp or pref PS filters into the nodes systems. Load Filter Node.
The program Zweistein http://www.zs4.net/ comes to mind where many artistic filters are used for Video Editing, including for VJay use.
If anyone has ever done the tedious task of rotoscoping, this sort of thing would be a welcome addition to Blenders toolset, be it in Compositing or Materials nodes or both.
m.a

#20

vote for frequency analysis node