Suggestions for future updates

Hey,

First of all, thanks for porting Fusion to OSX. That was my original #1 wish list item! I have a few suggestions, in case you're interested:

1) Pipes: I find them extremely messy, the way they always overlap. Once you do something that's beyond the extremely simple, you're bound to have your comp look like spaghetti. The square pipes setting doesn't help, it frequently makes connections less visible. I would love a system akin to what Cycling 74's Max/MSP is using (that's an audio app that uses "objects" very much like nodes), where you can shift-click on the background to make your own corner points, enabling, say, a pipe to go "around" your comp to the other side... it makes everything so much clearer.

2) Auto-save a-la Final Cut: That's OSX's system of saving just the first time, and then the app saves each and every step you make. If the app crashes (for me, Fusion crashes a lot, not just the 8 beta, also v7 on Boot Camp and Win8), you can reopen it and you are right where you left off. The system is elegant once you get used to it, you can save versions within the same finder file, and you can always duplicate and save with another name.

3) Documentation: At the moment, the docs are a mess. There are sections that are referenced and appear to be missing, stuff that's in there but not in the TOC, and plenty of things that could use some explaining. I would also like to see known quirks and their workarounds documented, maybe a couple of examples. Also, a keyboard shortcut could be used on every tool to bring up its reference page. Some more extensive docs on expressions with examples would be great as well.

4) Learning resources: I'm one of those people that just hates video tutorials, I'd rather have them written down. Maybe some common compositing scenarios could be documented, for example, how to set up a multipass EXR comp, or how examples of common particle systems. Also, guides on how to export data from popular applications would really help... this needs a lot of googling at the moment.

5) Linear workflow: This one's hard at the moment. I don't mind the setting up, actually I prefer to know what I'm doing rather than having the app do stuff automatically. But I'd want to be sure that, e.g. the renderer is working in the same colour space I have for the rest of the comp (there was a thread in this forum that got me really confused on that). Also, some tools don't appear to be working very well... I found it very difficult to use primatte, whereas converting to sRGB made it much easier, but then I could only use the matte on output, not the advanced features like despill etc. Also the despoil controls on the Matte Control do not work for me in linear, as well as the supression tab on the CC which just behaves strangely.

6) CC matching function: Maybe porting some code from Resolve, where you can have the CC set itself to match a background plate on another input? It doesn't have to be a one-click automagic thingy, just a good starting point that can be adjusted, to help blend, for example, 3D stuff with camera footage.

7) Camera solve and planar tracking: the more I can do inside Fusion, the happier I'll be... Also the point tracker could use some of Resolve's features, especially the "interactive mode" stuff. It's so easy to place tracking points that way.

8) Resolve-style qualifier curves: I figured out HSL qualifiers in Resolve with in about 10 seconds. I'm still struggling to do that easily in Fusion.

9) 3D path view: It could be useful to be able to view a path in a perspective-style view, with the common 3D navigation method, so that the Z-axis would represent time. Sort of combining the path view with the splines. I've never seen this, so it might be a stupid idea, just saying...

10) Insert calculation on animated parameters: This comes up a lot for me. Sometimes I'll animate something with some keyframes and edit the spline, only to later realise that I'd like to add, e.g., some noise to it, or an oscillation, or whatever, and then it becomes time-consuming. There could be a contextual menu item that "transforms" the animated control by inserting a calculation modifier, with the first operand already set to the original spline. Then you can add to it or whatever you want, and keep building the animation.

11) Lastly! Node-based interface for connections: The connections right now are accomplished with the contextual menus, published points, and expressions. But it could be beneficial to have them as nodes, for the same reasons that everybody loves about node-based compositors, compared to layer-based ones, or even more so, to text scripts. If you open an old comp, or someone else's comp, that's heavy with all sorts of connections, it can be hard to figure out. But the same thing could be accomplished in the flow, maybe on a different visual layer that can be turned on and off. So, a tool that is publishing a control could have an extra outlet, an "explode" data node could separate out the parameters needed by name, there could be data tools that do math and can combine many inputs, and a special "connector" tool could be used to specify the parameter than needs to change on its connected target node. This way, you can see and trace out the calculations. I know it feels that it can cause a lot of clutter, but if that layer can be turned on and off, it could be really useful. This is, again, a lot like Max/MSP works, and I love that.

Sorry for the long post, a lot of the stuff is wishful thinking, I know. I just thought I'd share. Fusion rocks already as it is.

Hi Dani,

irregardless of the validity of your other interesting points, there are some requests that are already fulfilled

1) Pipes:

try alt-clicking on a connection!

2) Auto-save a-la Final Cut:

Fusion has auto-saving but its interval is set to 10 minutes which is ridiculously high and you might have to redo a lot of stuff in case of a crash. Nuke saves 3 seconds after your last mouse click which means I hardly lose work at all.

6) CC matching function:

ColorCurves has a matching feature, CC has a histogram matching features and there's the white balance tool for manual matching. That being said, both techniques are really crude and I've never had any good results. I bet some of Resolve's features will bleed into Fusion sooner or later.

10) Insert calculation on animated parameters:

There already IS a context menu to insert modifiers on animated controls.

Hi Stefan, thanks for your message.

1) Thanks, I wish I knew that a week ago... Still, I think what I'm proposing is better yet. Think clicking once on the output, and then shift- or alt- clicking on any point in the background to make these connections as you go, with the final click on the destination. Unless you can already do that too and I haven't figured it out yet

2) I know about the existing auto-save... apple's method is something different altogether, and it combines auto-save with a versioning feature. It'll be an OSX thing by necessity, but maybe there could be an option? All the apple apps do that now, and after getting used to it, I find that it's really handy.

6) Agreed.

10) Yes there is and I can't figure out why I haven't seen it. I looked at the greyed out modify menu, and didn't even think to check the insert menu. Dumb. Thanks...

2) Auto-save a-la Final Cut: That's OSX's system of saving just the first time, and then the app saves each and every step you make. If the app crashes (for me, Fusion crashes a lot, not just the 8 beta, also v7 on Boot Camp and Win8), you can reopen it and you are right where you left off. The system is elegant once you get used to it, you can save versions within the same finder file, and you can always duplicate and save with another name.

Generation does something like this, but it's unworkable for Fusion.

1) You spend more time "doing things" in Fusion. Generation and FCP spend more time looking at things without changing them.

2) Comps in Fusion store the state of the comp, not a list of instructions. Fusionscript isn't even capable currently of recording everything you do in Fusion

3) Comps in Fusion are human readable. That's more valuable than fine grain autosaves.

4) Learning resources: I'm one of those people that just hates video tutorials, I'd rather have them written down. Maybe some common compositing scenarios could be documented, for example, how to set up a multipass EXR comp, or how examples of common particle systems. Also, guides on how to export data from popular applications would really help... this needs a lot of googling at the moment.

Agreed. A cookbook of common workflows would have two benefits:

1) Easier task-specific training. Assuming you were just given a task that you've never done before in Fusion, but it's not an unusual task for a professional compositor, having a resource that explains how to do things like frame repair, stereoscopic alignment and depth grading, greenscreen, CG integration, set extensions, etc. would be very useful. Heck, BMD could show off what their cameras could do by providing the footage. BMD could also afford to hire someone to make assets for this, which is a big barrier otherwise.

2) User stories. Not everyone working on making Fusion is a compositor. Rony has a lot of experience, but having task goals and user stories for Fusion's developers would expose any weaknesses and create test cases to develop against.

5) Linear workflow: This one's hard at the moment. I don't mind the setting up, actually I prefer to know what I'm doing rather than having the app do stuff automatically. But I'd want to be sure that, e.g. the renderer is working in the same colour space I have for the rest of the comp (there was a thread in this forum that got me really confused on that). Also, some tools don't appear to be working very well... I found it very difficult to use primatte, whereas converting to sRGB made it much easier, but then I could only use the matte on output, not the advanced features like despill etc. Also the despoil controls on the Matte Control do not work for me in linear, as well as the supression tab on the CC which just behaves strangely.

You can't set the colorspace of the 3D material tools any more than you can set the colorspace of a Merge. It's linear with abstract primaries. Sure, there are some tools that aught to respect different primaries, pretty much anything with luminance or saturation controls or anything that converts colorspace, like RGB to LAB and such, needs to read in the colorspace metadata.

Primatte doesn't work in linear space, dunno why. It's really frustrating as you have to convert to a mystery space that isn't in the documentation to get the math right. I should probably do some testing to figure out what at least the gamma is. I also agree about Matte Control. I hardly use it for that reason, preferring to use CCv or HCv. Not sure what you mean about the suppression tab in CC, though.

9) 3D path view: It could be useful to be able to view a path in a perspective-style view, with the common 3D navigation method, so that the Z-axis would represent time. Sort of combining the path view with the splines. I've never seen this, so it might be a stupid idea, just saying...

You can do that now if you want. Just set the Point values to be XY and the time to be Z and plot a path in 3D.

Code: Select all

{
   Tools = ordered() {
      Locator3D1 = Locator3D {
         CtrlWZoom = false,
         NameSet = true,
         Inputs = {
            ["Transform3DOp.Translate.X"] = Input {
               Value = -0.25,
               Expression = "Triangle1.Point1.X-.5",
            },
            ["Transform3DOp.Translate.Y"] = Input {
               Value = -0.421561501920223,
               Expression = "Triangle1.Point1.Y-.5",
            },
            ["Transform3DOp.Translate.Z"] = Input { Expression = "time*.01", },
            ["Transform3DOp.Rotation"] = Input { Value = 0, },
            ColorBlue = Input { Value = 0.05, },
            MakeRenderable = Input { Value = 1, },
            Width = Input { Value = 2048, },
            Height = Input { Value = 1556, },
         },
         ViewInfo = OperatorInfo { Pos = { 660, 82.5, }, },
      },
      Merge3D1 = Merge3D {
         Inputs = {
            ["Transform3DOp.Translate.Z"] = Input { Expression = "-time*.01", },
            SceneInput1 = Input {
               SourceOp = "Locator3D1",
               Source = "Output",
            },
            SceneInput2 = Input {
               SourceOp = "Locator3D2",
               Source = "Output",
            },
            SceneInput4 = Input {
               SourceOp = "Locator3D3",
               Source = "Output",
            },
         },
         ViewInfo = OperatorInfo { Pos = { 770, 115.5, }, },
      },
      Locator3D2 = Locator3D {
         NameSet = true,
         Inputs = {
            ["Transform3DOp.Translate.X"] = Input {
               Value = 0.224289510864764,
               Expression = "Triangle1.Point2.X-.5",
            },
            ["Transform3DOp.Translate.Y"] = Input {
               Value = -0.354707721620798,
               Expression = "Triangle1.Point2.Y-.5",
            },
            ["Transform3DOp.Translate.Z"] = Input { Expression = "time*.01", },
            ["Transform3DOp.Rotation"] = Input { Value = 0, },
            ColorBlue = Input { Value = 0, },
            MakeRenderable = Input { Value = 1, },
            Width = Input { Value = 2048, },
            Height = Input { Value = 1556, },
         },
         ViewInfo = OperatorInfo { Pos = { 660, 115.5, }, },
      },
      ImagePlane3D1 = ImagePlane3D {
         Inputs = {
            ["Transform3DOp.ScaleNest"] = Input { Value = 0, },
            ["SurfacePlaneInputs.ObjectID.ObjectID"] = Input { Value = 1, },
            MaterialInput = Input {
               SourceOp = "Triangle1",
               Source = "Mask",
            },
            ["MtlStdInputs.Diffuse.Opacity"] = Input { Value = 0.5, },
            ["MtlStdInputs.MaterialID"] = Input { Value = 1, },
         },
         ViewInfo = OperatorInfo { Pos = { 660, 49.5, }, },
      },
      Triangle1 = TriangleMask {
         CtrlWZoom = false,
         Inputs = {
            OutputSize = Input { Value = FuID { "Custom", }, },
            MaskWidth = Input { Value = 512, },
            MaskHeight = Input { Value = 512, },
            PixelAspect = Input { Value = { 1, 1, }, },
            ClippingMode = Input { Value = FuID { "None", }, },
            Point1 = Input {
               SourceOp = "Perturb1",
               Source = "Value",
            },
            Point2 = Input {
               SourceOp = "Perturb2",
               Source = "Value",
            },
            Point3 = Input {
               SourceOp = "Perturb3",
               Source = "Value",
            },
         },
         ViewInfo = OperatorInfo { Pos = { 550, 49.5, }, },
      },
      Perturb1 = PerturbPoint {
         CtrlWZoom = false,
         Inputs = {
            Value = Input { Value = { 0.25, 0.25, }, },
            XScale = Input { Value = 0.25, },
            YScale = Input { Value = 0.25, },
            Wobble = Input { Value = 8, },
         },
      },
      Perturb2 = PerturbPoint {
         CtrlWZoom = false,
         Inputs = {
            Value = Input { Value = { 0.75, 0.25, }, },
            XScale = Input { Value = 0.25, },
            YScale = Input { Value = 0.25, },
            RandomSeed = Input { Value = 10750, },
            Wobble = Input { Value = 9, },
         },
      },
      Perturb3 = PerturbPoint {
         CtrlWZoom = false,
         Inputs = {
            Value = Input { Value = { 0.5, 0.666, }, },
            XScale = Input { Value = 0.25, },
            YScale = Input { Value = 0.25, },
            RandomSeed = Input { Value = 24070, },
            Wobble = Input { Value = 10, },
         },
      },
      Locator3D3 = Locator3D {
         NameSet = true,
         Inputs = {
            ["Transform3DOp.Translate.X"] = Input {
               Value = -0.0220980448648334,
               Expression = "Triangle1.Point3.X-.5",
            },
            ["Transform3DOp.Translate.Y"] = Input {
               Value = 0.0168303984999657,
               Expression = "Triangle1.Point3.Y-.5",
            },
            ["Transform3DOp.Translate.Z"] = Input { Expression = "time*.01", },
            ["Transform3DOp.Rotation"] = Input { Value = 0, },
            ColorBlue = Input { Value = 0, },
            MakeRenderable = Input { Value = 1, },
            Width = Input { Value = 2048, },
            Height = Input { Value = 1556, },
         },
         ViewInfo = OperatorInfo { Pos = { 660, 148.5, }, },
      },
      Merge3D1_1 = Merge3D {
         CtrlWZoom = false,
         Inputs = {
            SceneInput1 = Input {
               SourceOp = "Merge3D1",
               Source = "Output",
            },
            SceneInput2 = Input {
               SourceOp = "ImagePlane3D1",
               Source = "Output",
            },
         },
         ViewInfo = OperatorInfo { Pos = { 880, 115.5, }, },
      },
   },
   ActiveTool = "Merge3D1_1",
}

11) Lastly! Node-based interface for connections: The connections right now are accomplished with the contextual menus, published points, and expressions. But it could be beneficial to have them as nodes, for the same reasons that everybody loves about node-based compositors, compared to layer-based ones, or even more so, to text scripts.

Agreed. You can accomplish this with Fuses and Plugins easily enough. For things like Numbers and Points it's really simple, but Gradients and Gridwarp Meshes are trickier.

Hi Stefan, thanks for your message.
10) Yes there is and I can't figure out why I haven't seen it. I looked at the greyed out modify menu, and didn't even think to check the insert menu. Dumb. Thanks...

I'm gonna use you as an example for why I think the context menus need to be cleaned up and need to be smarter instead of just graying out all the inapplicable commands

Thanks for your suggestions Dani. We do read everything here but don't always get time to reply. We'll certainly look at incorporating any good ideas we can in the future.
cheers

Hey,

5) Linear workflow: This one's hard at the moment. I don't mind the setting up, actually I prefer to know what I'm doing rather than having the app do stuff automatically. But I'd want to be sure that, e.g. the renderer is working in the same colour space I have for the rest of the comp (there was a thread in this forum that got me really confused on that). Also, some tools don't appear to be working very well... I found it very difficult to use primatte, whereas converting to sRGB made it much easier, but then I could only use the matte on output, not the advanced features like despill etc. Also the despoil controls on the Matte Control do not work for me in linear, as well as the supression tab on the CC which just behaves strangely.

Hey Dani,
I was using the De-spill in the Matte Control, just the other day, as well as the Hue curves and UltraKeyer. The Matte Controls, worked just fine in Linear for me. Perhaps I am missing something, perhaps you could expand on this?

Actually, just to confirm, are you talking about the ability to select colours? I have to admit for fine control I do not use this tool. Just want to know exactly what your issue is...

cheers

I was using the De-spill in the Matte Control, just the other day, as well as the Hue curves and UltraKeyer. The Matte Controls, worked just fine in Linear for me. Perhaps I am missing something, perhaps you could expand on this?

Actually, just to confirm, are you talking about the ability to select colours? I have to admit for fine control I do not use this tool. Just want to know exactly what your issue is...
cheers

Basically I couldn't get it to despill enough. The colour was not pure green, though, from a proper greenscreen. Still, I think in srgb it worked better.

Let me also add that the ranges in the CC are very hard to define to get actual shadow, mid tone, and highlight corrections.

Basically I couldn't get it to despill enough. The colour was not pure green, though, from a proper greenscreen. Still, I think in srgb it worked better.

Let me also add that the ranges in the CC are very hard to define to get actual shadow, mid tone, and highlight corrections.

Hey Dani,

Right, this is why I asked if you had a problem because you could not select the colour. It sounds like you should use a different tool. Try Hue Curves, or some other technique. The green being selected by the tool is specific to deal with the most likely green screen colour. this way it will not effect other hues of green. It is a bit of a blunt instrument. However I have had good success with it. Going to sRGB only changed the Gamut, which is the same as Rec709 except the gamma is different so you are just changing the way the colour values are distributed in relation to your viewing environment. However, are you changing from linear light to sRGB in the comp? This is taking you out of the linear light world and into a non linear gamma representation and if you are picking green it may offer some interesting results, but in my opinion this will not give you the results you are wanting.

Can you expand on the issue. Again I am using this tool with no real issue for CC work. Are you trying to pull some kind of key? You can get keying results for sure. Just not what you may want? I can get really good keys if I want from the ranges. In fact it acts just like the luma keyer but with a bit more visual feedback as to what you are doing. But basically I can get the same thing. Can you explain what you are wanting?

cheers

Let me also add that the ranges in the CC are very hard to define to get actual shadow, mid tone, and highlight corrections.

That's a common issue. The ranges are linear, but your expectations of what the ranges should be is non-linear. It would be nice if there was better controls for making those selections.

Can you expand on the issue

Hey Mark,

Thanks for your interest in this. I needed to do screen replacements, where the actress would pass in front of the screen all the time, in multiple shots. There was hair and tons of motion blur, since this is a dance video. On set the monitor was covered with green paper. We worked in linear because there's also a lot of 3D renders from Vray to be composited elsewhere. There was also tracking to deal with, so at least for me, being a noob, it got pretty complex. It didn't help that the green backing was lit very inconsistently.

I think you are right that this is more of a colour issue after all... What I ended up doing, since Primatte misbehaved in linear, was to convert the footage to srgb and CC it so that I maximise the separation of the screen, just to feed the keyer. Then I'd just use the alpha channel from Primatte and composite outside. However, I got a lot of green fringes and green tints in semi-transparent areas (in hair and motion blurred parts, also the occasional fringe). I was hoping to kill it easily, since it's A LOT of work to do it manually, and I couldn't use Primatte's own tools. In the end, I did some math with the masks, or edge detection, blurring, etc, and used it as a mask for the CC and it worked. It just took forever, and in a few cases I had to help the keyers with some roto work with the paint tool. Surely there was a better way, but I'm learning.

besides going into sRGB it sound like a "normal" keying workflow. building you matte and use that on the footage in a seperate tree/branch of the flow. doing some roto is also very common.

i would probaly doing the despill on the footage (before applying the matte) and the combine it with the matte you create with the keyer and then merge it over the background

most of the times i also create different keys for different parts of the footage and combine them into one matte

i would probaly doing the despill on the footage (before applying the matte)

I despilled after applying the matte, but before I merged with the background obviously. Is there a difference from what you suggest, like when the matte gets multiplied? I'm a bit unclear on that...

The different parts technique I've seen done a lot in videos, but my material has too much motion for me to be able to pull that off... the camera moves, so I'm tracking the replacement area and also using the tracker's centre to connect to the pivot of a couple of transforms that I use to scale the image a bit for reflection simulations/displacements/light wrap thingies. I also use the track for animating the garbage matte and a spline of the original (unoccluded) shape. If I have to track the moving objects (persons!) on top of that I'll go insane! But on a couple of occasions I had two different keys being pulled and selected one or the other depending on what part of the body was occluding the replacement area (say, hair or arms etc), by animating a dissolve. Sometimes I find it quicker to use the paint tool...

for me the priority is on
- multipass exr et custom channel creations like nuke, workflow is actually wayyyyyy to slow
- Color managment / ACES support and a real and clear linear workflow for all files setup
- nodegraph organisations : really strange to have input / output moving.
- better render 3D node : the defocus is not that nice (or i don't use is well)
- mask nodes can be merged, like in nuke the roto / rotopaint node. actual nodes are a slow workflow
- 2D and 3D tracker : copy nuke ones, those are great
- a node like shuffle copy in nuke, the bolean node is great but miss some features in front of nuke one
- expression node : to have custom expression to create what whe need, like custom despill for example

- better render 3D node : the defocus is not that nice (or i don't use is well)

- expression node : to have custom expression to create what whe need, like custom despill for example

I think there are more recent threads for feature requests. Last post in this one was two years ago.

For your depth of field in the Renderer3D, try turning quality in the Accumulation Effects controls up way beyond 10. I'm not sure why that's the default max scale, but you can always access larger numbers by typing in a numeric field. DoF can look quite nice with values of 40 - 80, or even higher if you're doing something really shallow.

The expression node you're looking for is called the CustomTool. The Tool Reference document has some good examples of how to use it. I've also got a blog post covering Expressions that has some more samples at http://www.bryanray.name/wordpress/blac ... pressions/ Scroll down about half way to the header "Per-Pixel Expressions with the Custom Tool."

There are also CustomVertex3D, pCustom, and pCustomForce. The first gives you access to properties of 3D objects on a per-vertex basis, and the other two are for particles. I did some fun stuff with pCustom in this article: http://www.bryanray.name/wordpress/link ... ic-fusion/