[prajjwal]

@7even 29429 wrote:

… and what was the conclusion?

As with everything, it depends what you need the software to do.

Nuke excels at VFX work where a comp team will work on a selected VFX shots for days/weeks/months.
Nuke is really at home working with CG render passes, especially .exr files.
It has superb customisation support through Python and TCL scripting, gizmos (macros in Shake-speak) and the NDK (Nuke SDK for plug-in development).
It is optimised for wall-to-wall 32-bit float “linear light” support.
Full support for RGBA (and more!) channels throughout the app.
Elegant tools for stereoscopic workflows.
Optimised for good interactivity even on modest hardware through the viewer only rendering what you see (either rendering a region of interest when zoomed-in or line skipping when zoomed-out).
Since Nuke is supported on all 3 major platforms (Win, Mac, Linux) on various hardware, it’s quite easy to leverage existing 3D render farms to support Nuke.
Bigger VFX shops seem to be all moving to Nuke (ILM, Digital Domain, SPI, Framestore, Animal Logic, MPC, etc…)

Now that Smoke Advanced has a batch module, I really don’t know who Flame is targeted at. It would seem that Smoke is the new flavour of the month for advertising/client driven/lots of tape IO sessions. And I honestly haven’t heard anybody getting Flames for film VFX work in a while now.

Anyways, Smoke (or Flame) excels at one-man-band client-driven sessions that involve lots of tape IO (as in the client walks in with a stack of HDCAM tapes and expects to walk out with tapes for 4x 30-second spots, ready-to-air, in 2 days).

Since Smoke is really an editing/finishing tool, it obviously blows Nuke (non-existent) editing tools out of the water.
Smoke has a great disk-based clip player. No cacheing to RAM necessary to play long clips like FrameCycler.
It has a 2d tracker that’s very efficient at locking onto tricky targets.
The color correction tools like Color Warper feel a little bit more polished than Nuke’s.
The interface is really wacom-friendly with big buttons and minimal use of right-clicks and double-clicks.
Seasoned editors seem to like the way the app is laid out and the one-stop shop mentality of Smoke / Flame.

–Xavier

[prajjwal]

Hi,

I was a FFI user for years and now a Nuke guy.

I do a lot inside the “colour pot”. (The little colour pizza next to most controls).

That way if you enlarge the control window that pops up, you get nice long sliders to play with.

Another bonus is that you can adjust, say, RGB Gain or RGB offset but with HSV controls if you want. I often crank up the saturation, play with the hue until I get the right tint, then dial the saturation down.

It’s really nice to be able to do micro tweaks to all 3 channels by only playing with the hue or saturation.

What you might be experiencing too is the switch to linear color space. “Mid-gray” is 0.18 instead of 0.5 in gamma-2.2 world. This means that curves and ranges are squeezed to the “left”. It was suggested to Foundry to offer the option of non-linear display to remap 0.18 to the middle of the range in the GUI. I don’t know if they’re going to do it.

So overall, the GUI of Nuke is not as slick as FFI, but you will get used to it pretty quickly.

Colour warper and 3d keyer are really the only things I kinda miss from FFI. But I wouldn’t trade alpha channel support in all nodes for these 2 tools. 🙂

— X

[prajjwal]

@andy_dill 28113 wrote:

but I have a hard time with the idea that what we all do for a living is even remotely similar to what Picasso, Michelangelo or Klimt was up to.

You mean that doing a pack shot on a box of laundry detergent is not art!?

🙂

— X

[prajjwal]

TRSEFX,

What you’re looking for is the “Card3D” node in the Transform menu (not to be confused with the “Card” node from the Geometry menu).

Ignore the “cam” and “axis” inputs. You don’t need them if you want to do a simple 3D transform like an “image” in Action. (You don’t need a scanlineRender either).

Make sure your output format is set properly, otherwise your image might get cropped.

— X

[prajjwal]

Well, if Toxik is to gain any traction, they have to “steal” users from other packages.

Now my question for Toxik users: what are the top 5 reasons for switching from Nuke to Toxik?

— Xavier

[prajjwal]

Sapphire ScanLines?

— X

[prajjwal]

Che,

I have worked in both Flame and Shake for various types of effects on feature films (not commercials or music videos).

@CheGuevarra 24595 wrote:

1. Flame vs Shake Paint

Shake paint is much more limited than Flame paint, but it is vectorial which means that you can replace what you are painting on without losing your paint strokes. You can paint colors, clone, reveal and smear (which NOT as nice as the “drag” brush in Flame). That’s it.

However, if you are creative, you can simulate dodge, burn, wash etc… by painting coloured strokes on black (or 50% gray) and using transfer modes like “softlight” to blend your strokes on top of your image.

The MAJOR pain with Shake’s paint, is being unable to quickly reveal pixels from other frames in the sequence (like in Flame, go to frame 60, hit “save”, back to frame 5, reveal “saved”).

The paint in shake is one of the few things that actually use the graphics card, so make sure to have a “nice” card if you want to have good interactivity.

Verdict: Flame’s paint is much more refined. For easy to medium paint jobs, Shake works good enough. For hard core background reconstruction, etc… Flame still has the edge. As someone mentionned too, Combustion’s paint is supposedly pretty nice, although I never worked on it.

@CheGuevarra 24595 wrote:

2. Flame Keyer vs Shake

Out of the box, flame comes with 2 types of 3d keyers (3d keyer and Master Keyer), 2 types of “chroma” keyers (the “old” Keyer and the diamond keyer) and the “difference” keyer. They are all excellent at what they do. However, Flame doesn’t come with a color difference keyer, so investing in a 3rd party plug-in like Keylight might be a good idea. You could also build your own in batch.

Shake comes with 1 3d keyer (Primatte) and 1 color difference keyer (Keylight) and you can use the ISubA node to do difference keys. The “ChromaKey” node in Shake is pretty lame in my opinion, so I made a macro to emulate the Flame keyer. You can download it (among other things) from my site: http://www.pixelfudger.com. To have really nice spill suppression, check out this macro:
http://www.fxshare.com/shake/downloads/macros/keying/SpillReplace-4710.html?tab=Description&&id=4710

If you need to extract stuff from blue/green screens, Keylight is usually pretty good. To extract more “arbitrary” colors, Primatte usually handles it better.

I am a big fan of Flame’s 3d keyer, but when you know how to use them, Primatte and Keylight can give you equally good results.

@CheGuevarra 24595 wrote:

3. Flame Roto vs Shake

As someone else mentionned, tying up a Flame to do hardcore roto is kind of a waste of money. However, if you need to compare both software in a vacuum where money is no object and other software doesn’t exist, I would say roto is generally nicer and quicker in Flame.

The biggest problem in Shake is the absence of the “shape” animation curve. This makes editing animation really annoying because you can’t slip individual keyframes easily. Even worse, lets say you have 2 shapes within a single RotoShape node. If both shapes have a keyframe at frame “5”, it is impossible to delete that keyframe only on one of the 2 shapes. This virtually forces the user to stick to 1 shape per RotoShape node.

Also, if you have roto that went “wrong” from frame 50 to 100, it is impossible to multiple select shape keyframes to delete them. If you had 1 keyframe per frame, you would need to hit the “delete keyframe” button 50 times in that example.

Shake does have “tracer” style per-vertex softness, but it lacks a global softness control. This means you will likely end up with a blur node right below a lot of your RotoShape nodes. Fear not though, because Shake is MUCH more efficient than Flame for blurs though, especially since rotoshapes generate a proper DOD. This means that if you have a tiny shape , only a few pixels get blurred instead of the whole frame.

@CheGuevarra 24595 wrote:

4. Flame Render Speed vs Shake.

You can run Shake and Flame on almost identical hardware these days… With either software, if you do complex stuff, you will want to send renders on a render farm. If you MUST render locally, it is very easy in Shake to render locally in the background while you work in the foreground. If you have a multi-core machine, you can limit how many cores are used by the render, so you still have a bit of CPU power to keep working.

Burn licenses (Flame render farm) are ridiculously expensive… even more so if you use Sapphire sparks (who doesn’t?). On the other hand, Shake render licenses on Mac are free (except for plugins, but usually much cheaper than the FFI version). I’ve always tought that a stack of Mac minis would make a cool poor man’s render farm, but if you have the money go with xServes.

By default, Flames come with very fast local storage (the Stone). That is the “secret sauce” to Flame’s speed. Big shops often have Shake sip the images from the network for ease of media management (easier to keep track of stuff, easier to bounce shots from artist to artist). If you are a small team (or solo), DO setup your shake workstation with fast storage and put your images on there. You can thank me later. 🙂

My experience has been that Shake is much more optimized in it’s rendering, but since Flame runs on linux/intel/64bits, both software can be pretty quick.

Do not waste your time on older non Intel Flames though. An octo-core Intel box (running Shake or Flame) will eat SGIs for breakfast. Plus, getting a good SGI sysadmin can be a challenge these days.

@CheGuevarra 24595 wrote:

If I want to dedicate myself 90% to a certain program, should it be Flame or Shake.

It depends what kind of artist you want to be. Feature film or commercials? Client driven one-man band sessions or working in a larger team? Flame is still pretty big in commercials, but Shake has the upper hand in features. My advice, learn both. 🙂 I am very glad I did. If you started with Flame, learning Shake (or Fusion, or Nuke) will seem unnatural at first, but trust me, it is worth it. Once you are comfortable in Shake, learning Nuke or Fusion if you need to will be a snap.

Other random notes for the Flame refugee on Shake.
1- Get Framecycler. The Shake flipbook just doesn’t cut it.
2- Get Sapphire plug-ins. You will feel more at home.
3- The coolest stuff in Shake can be found in macros, not built-in tools. Browse the stuff at http://www.fxshare.com.
4- Wrap your head around RGBA. The “matte” input should NOT be used to comp layers on top of each other… that’s what the alpha channel is for. This is probably the #1 mind-fudge for Flame expats.
5- Wrap your head around floating point. Negative colors and “brighter-than-white” colors are possible and can be extremely useful… but can cause trouble if you don’t know how to use it. This is the #2 mind-fudge for Flame expats.
6- Change your “pan” hotkey from “space-bar” to “middle pen button” in Flame, then sit on your non-pen hand for a day. 🙂 That way, when you bounce between boxes, you won’t feel disoriented as much.

Good luck.

– Xavier

[prajjwal]

@jimmybee500 24547 wrote:

I will have a good look at Nuke…I did watch a few tutes on it a while ago and the interface looked horrible and very clunky, but now The Foundry has it I’m sure they’ll do a sterling job with it. I may buy Shake in the meantime then…only thing is Shake doesn’t have a true 3D environment and Fusion’s true 3D scene import/3D particles, primitives and text looks very powerful and I really did think Fusion was becoming the cheap seat Flame backup tool of choice. Seems not the case…

Yes Nuke’s current interface IS absolutely horrible. No doubt about it. I’ve personally seen a demo of a beta (or alpha?) version of the next Nuke interface. The Foundry completely reworked it. Much much better. (You can get a glimse of it on one of this year’s NAB video podcasts) Nuke does have a decent 3d environment with .obj import textures, projectors, bicubics and bilinears. No native 3d text or particles though.

Fusion does have some advantages, they just don’t apply to pure feature film compositing. If you do a mix of comp, motion graphics, titles, etc… then Fusion might be a better choice for you because of the particles, 3d text and AE plug-in support. If you are a Flame op, you will also get a warm fuzzy feeling when you open Fusion’s color corrector. (To get the same fuzzy feelings in Shake, check out my macros on http://www.pixelfudger.com)

Back to the original question, Fusion should run decently on a Mac booted in Windows (via Boot Camp). I wouldn’t count on VMWare or Parallels though. I think all the drivers that come with Boot Camp (to drive all the Apple hardware within windows) are only 32 bits, so forget about Fusion64.

— Xavier

[prajjwal]

@Imhotep397 24541 wrote:

It’s depends on how quickly you need your compositing solution. A combination of Shake and Motion or even just Motion and the Conduit plug-in from DV Garage out match Fusion, feature wise, pretty easily but “Phenomenon” will be out in the next couple of months. Nuke imo is also better solotion for 99.9% of tasks than Fusion and it also runs natively under OS X. Windows under bootcamp works exactly like any other PC, but it’s and inconvenient option to have to partition and reboot just to get into Windows. If for whatever reason you’re still stuck on Fusion you’d be better off using VMWare Fusion under OSX and you should be able to run Eyeon Fusion in Windows inside of a Window, 64-bit if necessary, with no performance hit and when you get the blue screen of death or everything freezes you can just close the window and continue working on OS X.

Have you tried Digital Fusion under VMWare Fusion?

Last time I tried, it failed. I’m guessing because of the lack of proper OpenGL drivers. Maybe things are different using the very latest version of VMWare. Digital Fusion booted just fine in Boot Camp though.

I wouldn’t seriously recommend using your main compositing app inside a virtualisation software anyways… running two OSes on top of each other is OK for using Outlook and PowerPoint, but not really efficient for RAM and disk IO hungry operations like compositing.

My suggestion is give Shake a spin. At 500$, you can’t really go wrong. Shake, Fusion and Nuke all work on the same basic node-based architectures anyways, so if you want to switch to another package down the road, you don’t have to relearn *everything* (unlike if you move from Flame to AE for instance). And there are *tons* of resources to learn Shake (books, DVDs, online, etc…)

Regarding “Phenomenon”… my bet is that it is pure vapourware. If they had a really kick ass high end compositing package (a true successor to Shake), half of the key Shake staff wouldn’t have quit to go work at The Foundry (makers of Nuke and Furnace) like they did. They would have stuck around to be part of this new revolutionary product, wouldn’t they? Also, how long does it take to come up with a comp package when you already have the code for Shake, Motion, Final Cut, QuickTime, Aperture and the whole friggin OS? Wouldn’t it be in beta already? It hurts to say this, but Apple has bigger fish to fry than a tiny little niche market like compositing. They said at NAB that Final Cut is in the 800 000 users range. They are about to sell their 5 millionth iPhone. How many compositors in the *world*? 2000 maybe? Considering that big studios like ILM, SPI, DD, MPC, etc… are very unlikely to switch to Mac, how big a market is compositing for Apple? Tiny. Don’t hold your breath for “Phenomenon”.

— Xavier

[prajjwal]

Maybe make sure your DPX files are plain jane 10bit RGB — not RGBA, or some other bit depth?

— Xavier

[prajjwal]

@rohit 23467 wrote:

increasing the exposure by one F-stop results in an image with the luminance value double that
of the original….does this help ?

Actually, increasing by one F-Stop doubles the luminance of the original IN LINEAR color space. If you are working in FFI, there is a 99% chance that you are working either in video space (aka sRGB or gamma 2.2) or in log space.

For video space (HD, DigiBeta, some JPG you found online, etc…) do a gamma of 0.45, then gain 200%, then gamma 2.2. I suggest using 3 separate CC nodes in Batch, as I am not sure of the order of corrections within a single CC node. (You should be in 12 bits, otherwise the banding is gonna sting!).

Actually, even in 12 bits, this is pretty brutal, so I suggest taking a peek and emulating the result with a curve.

For log space, take your original Cineon file, add a LUT Editor with the default values (Log to Lin, ref white 685, ref black 95, highlight 4095, shadow 0, Gam Film 1.0, Film Gamma 0.6, Softclip 0). Add a CC node after with a gain of 200% then apply a LUT Editor with default values but this time Lin to Log.

Again, I strongly advise to take a peek, then emulate the look with a curve in CC.

**SIGH** Now if FFI could give us 32 float or even 16 integer bit depths… we could finally join the world and work in linear like everybody else… Then working in f-stops would only require a single gain adjustment…

— Xavier

[prajjwal]

Getting a pass without CC is a great idea. However, if the original un-CCd material is still too grainy, and if you don’t mind a side trip to After Effects, this plug-in has done wonders for me in the past:

http://www.neatvideo.com/

— Xavier

[prajjwal]

Dear Lazy,

😉

To do a screengrab of the entire screen, in a shell type:

scrsave myimage.sgi

Just make sure your shell is not covering the bit you want to capture.

This will create the file myimage.sgi in the current directory.

— Xavier

[prajjwal]

shidarin wrote:

•The default-preincluded sparks have been upgraded to a new level of uselessness with fun sparks like “invert”

The sparks that discreet provides are meant as tutorials for people who want to program their own sparks.

The “invert” spark was quite useful to me when I wanted to understand the spark API.

That being said, I agree that most of them are pretty useless to me as an artist. 🙂

— Xavier

[prajjwal]

Hi Pete,

You can use the Timing tab of the fileIn to emulate average.

Simply set the values like this:

reTiming: speed
speed: 1
retime mode: Blend
retime bytes: I recommend 16 or float
weight: 0
range: 3

This will give you a 3 frame average (past and future). Play with range to average over more or less frames.

To compound a 30 frame clip into a single frame, set the range to 30 and render frame 15.

This should get you close to what you’d expect from FFI.

Good luck.

— Xavier

[Author]

Posts