Open source Softimage!

[Hirazi Blue]

As Blender has been gobbling up all interesting developments in OSS 3D over the last couple of years, moulding them into Blender's way of doing things (in a rather Autodeskian fashion), I respectfully disagree. Blender isn't merely an open source 3d project, it has de facto become the only one. And monopolies, even of the open source variety, are bad in my book.

[BitPuffin]

Healthy competition is extremely valuable in the open source world as well. You definitely do need it. Take for instance btrfs vs ZFS. BSDs vs Linux. The BSDs versus each other, the linux distributions vs each other. X11 vs Mir vs Wayland. There is lots of instances where some healthy competition has been nothing but good.

I have many reasons not to work with Blender. One is an ideological (and probably somewhat childish in many eyes) reason - GPL.
But the main one is that the reason I'm doing this is to learn how to build 3D tools. I could learn things from studying Blender's source code but
a) I really don't like reading other people's code (who does?)
b) The knowledge doesn't stick nearly as well as when you actually do it and figure things out on your own.

Another thing is that I wanted this product to feel comfortable for SoftImage users. I don't think Blender (or any of the proprietary alternatives) feel much like softimage. And it's not like SoftImage users are too dumb to learn those tools, but I think they were attracted to SoftImage for a reason. As I'm learning SoftImage I try to figure out what

EDIT:

Another thing, Blender has a reeeeeeeeeeaaaaaaally long history. This is cool. But that probably means there is some stinky parts in Blender's architecture that are hard to remedy (just a suspicion, I don't have any proof of this). I wanted to experiment in this project to try and write it with a very modern programming style using modern language features etc. Mainly because I'm curious as to how well these work out!

[csaez]

Healthy competition is extremely valuable in the open source world as well. You definitely do need it. Take for instance btrfs vs ZFS. BSDs vs Linux. The BSDs versus each other, the linux distributions vs each other. X11 vs Mir vs Wayland. There is lots of instances where some healthy competition has been nothing but good.

Wait a sec... BSD isn't open source (maybe you're refering to freeBSD?), in the other hand all Linux distribution are based in 5 main distros and there's a lot of forking in the mix, at the end, the "competition" has to do with a matter of taste when chosing the different components of the system (that's why I love arch linux). Wayland/Mir are not a competitors to X11 but its sucesor... and it has been an extremely controversial topic, the same was true for upstart vs systemd (no many FOSS people like Canonical approach).

I think you are missunderstanding the meaning of open source.


Anyways... good luck with the project.

[FXDude]

Healthy competition is extremely valuable in the open source world as well. You definitely do need it. Take for instance btrfs vs ZFS. BSDs vs Linux. The BSDs versus each other, the linux distributions vs each other. X11 vs Mir vs Wayland. There is lots of instances where some healthy competition has been nothing but good.

Indeed, we all saw that in recent developments of game engines.

Otherwise for your project, if not using Blender code, you would have to find some shortcuts by using some premade libraries,

Just having talked talking about the wish for a Fabric DCC, with Paul making it clear that perhaps a 3rd party could do it but not them,
so perhaps you could be that 3rd party?

I wonder how many worked on Natron (the Nuke open source Clone)

[BitPuffin]

Wait a sec... BSD isn't open source (maybe you're refering to freeBSD?)

BSD (at least in their community) is well understood as referring to the different open source ones, as they are the only ones that are still active. Such as FreeBSD, OpenBSD, NetBSD, DragonflyBSD.

About the distros you are ignoring a lot of the smaller distros that aren't based on one of the main ones (such as crux or morpheus). Regardless, even if there were only 2 "main" bases these would still be competing. The main ones have competing package formats and managers and they are not just different in taste but also have a lot of distro-specific software packages that compete (which could be integrated in the other distro too if they wanted of course). If you want another clear example of healthy competition in the open source world you have GCC vs LLVM/Clang which has caused GCC to improve vastly the last years and Clang to improve a lot in response to that. You also ignored a lot of the other examples I presented, maybe you didn't know what they are? A lot of the Wayland devs are working on X11 sure, but Mir is competition to wayland. I omitted upstart and systemd as even canonical has given up on upstart so it's not really relevant.

The difference is the open source world is that sometimes the competitors can borrow code or at least use each other's implementations of the best idea as reference.

I think you are missunderstanding the meaning of open source.

Please

Anyways... good luck with the project.

Cheers mate!

Otherwise for your project, if not using Blender code, you would have to find some shortcuts by using some premade libraries,

Yeah perhaps! So far I'm purposefully avoiding library dependencies because I want to try things first (mainly so that I'm not juggling things that I don't know how they work) before I integrate a premade solution. But I can certainly see myself probably integrating some libraries. For example Bullet for physics simulation or Cycles (blender's Apache licensed Path Tracing renderer) to fill in some gaps. At the very least as placeholders and/or options!

Just having talked talking about the wish for a Fabric DCC, with Paul making it clear that perhaps a 3rd party could do it but not them,
so perhaps you could be that 3rd party?

Don't know! As far as I can tell Fabric isn't open source so I would be cautious to spend my effort with it. I'd love to be proven wrong though .

Either way, as a reminder: as I mentioned in my initial post I'm not sure how far I will take this as it is just research/playing around to me. What I do want to do is to remind people that starting these kind of things are feasible and you can control and build the tools you want to use. You'll get there much faster if you start doing it rather than thinking about if it's doable or not. Of course it is! Or it wouldn't have existed in the first place. So if the only thing I achieve is inspire someone to work on a softimage application (or any kind of DCC) then I consider it a huge success!

Cheers everyone

[Hirazi Blue]

One thing I wanted to add to the mix is that to me Softimage perfectly shows that a well-designed workflow trumps a rich feature set anytime. Nothing in Softimage, not even ICE, is as important as this seemingly effortless workflow between its constituent parts. Having looked at Maya, its feature set is way more impressive, but its workflow is ridiculous IMHO. That's why I returned to Softimage swiftly after my failed attempts at using Maya.

Oh and not to invoke the wrath of the Maya crowd: very much IMHO, obviously.... ;)

[Draise]

Interesting debate. I do think another strong Opensource 3D program would be more than healthy for everyone - Blender included. I would love to see a full blown DCC that can splice into anything through Fabric Engine - but yes.. the basis of it wouldn't have been built by you and it may not be opensource.. thought 50 free seats certainly might make up for that concept.

Though.. Fabric with GPU accelerated Python based KL C++ compiled code and visual programming would be awesome to work with others - also being something agnostic to where it's applied - incoporating into any workflow and software.... just a thought.



Things I like about SI:

Unlimited Floating Panels - refreshable and lockable (surprisingly very handy - wish Blender had this..)
Middle mouse click on tool root to repeat action
Tearable menu's
Context sensitive Menu groups or tabs (I know it's cliche and everyone does it.. but it does make things handy when you are feature rich).
Option to create visual buttons and menus (Wishing there were more to choose from in SI tbh)
Dynamic Context based all inclusive Explorer/Outliner (IMPORTANT).
Schematic View (the simple one it has is handy)
L Click anywhere in the viewport and it'll control the widget regardless (don't have to use the widget nor annoying middle mouse button! - surprisingly handy)
Right click in viewport to change navigation properties (wish they also included axis restriction and planes, to click and drag anywhere without the widget to move such thing, to normal, world or local axis - a la trueSpace style)
Sticky keys (press and hold to activate shortcut, release to return to previous tool)
The M tool with proportional editing
Everything can be animatable with the little keyframe button beside it - including the keyframe button in the panel.
Animation layers, NLE mixer (with clip groups and blending), dopesheet and Fcurve editor - I think for animation it's very standard and necessary.
Procedural Operators and Freezing workflow - Non Linear modeling and animation workflow (IMPORTANT)
Ability to build secondary interfaces for rigging, animation or tools either with ICE or other parameters and expressions
Per object viewport and render display attributes (that can be overriden and animated)
Render passes and overrides and partitions
Enveloping
Project file structures
Anything can be used for enveloping - nulls, curves, objects
Node based shaders and easy material library
ICE particles and strands
Syflex, Crowds and Lagoa in ICE
Local and World animation and transform attributes
Exocortex ICE compounds
the Scatter ICE compounds
Other procedural ICE compounds, that don't come directly with SI, but certainly kick ass.


I would also like to have a Schematic view have it all, in a way, with access to the operators, hierachy, ICE programming nodes, simulation nodes, animation attributes, etc - with display filters. An ICE like system , Shader, Hierachy, Shape, etc. trueSpace had a cool one, where everything was in nodes, not just in an outliner/explorer - but it had no filters (so you had to dig deep to find the shaders, for e.g.). It wan on Javascript (so you could build your own nodes) and.. it was slow and clunky, but a great concept. If you had a dynamic schematic view with something like ICE that has access to everything, color coded, and with display filters - that'd be nice. (objects only, shaders only, render attributes, rigging attributes, all, etc). You could even do image processing in the schematic node tree.

You get my drift.. but that is just dreaming.

Blender is too fractured in different arenas when it comes to nodes... just one place for everything in the 3D world would be nice.

Ok, this isn't a feature request is it? More about Softimage?

Ok, well I'm in love with both trueSpace and Softimage - both now dead, and still find it hard to get into Blender or anything else, mainly because of the above things.

Great incentive, very ambitious, hope something can start or even strengthen with what is there - even if it's just a simple blank slate with the ability for others to visually program and create their interfaces with a cool 3D viewport that runs on DirectX11 and 12, and the simple tools of SI and the potential of it growing bigger and more useful.. that'd be cool.

[caledonian_tartan]

...I think they were attracted to SoftImage for a reason. As I'm learning SoftImage I try to figure out what.

General Efficiency



the first thing i learned with softimage, was not "to make a cool tool work", but to make a great tool work in its best performance and user efficiency.

That's what makes it impossible to create a DCC from scratch.

there are more than 20 years of professional experience inside our beauty!

maybe having a fabric core with an open source user based UI could be pretty nice ;)

[BitPuffin]

One thing I wanted to add to the mix is that to me Softimage ... etc

Yeah I think that's a good point. A lot of tools out there overlap in terms of what they can do. But it's the ones that are well designed that let you be really productive


Draise: Thanks for your excellent description of features you like etc! I will try to look up these things in softimage so that I can fully see what you mean . In regards to the schematic view that let's you dig in to it all I certainly think that would be very possible. I don't think I would store things like that in memory for efficency and CPU cache hotness etc. But combining all of the different types of things in to one view that you can filter doesn't sound too unrealistic if you ask me. Yeah, a blank slate that you can manipulate might be all it turns in to. I don't know yet how much I'm gonna spend on this! I just think it would be nice to get the ball rolling ;;).

caledonian_tartan: Well, I still don't think that makes it impossible. Maybe if there wasn't anything you could use as reference for what's been proven to work efficiently then yeah it will take time to nail it down. But when there is something like softimage that you can draw inspiration from in how they integrated their tools in a nice workflow I think you can build it faster than they did. Since they already did all of the grunt work. Regarding fabric engine that's gonna be a no unless they open source it. I don't think the solution to having an unreliable (in terms of will the product survive?) closed base (softimage), is to move to another one. Maybe instead what it would turn in to is something like fabric engine. Well. That's actually probably more like what I'm gonna build when I have some experience building 3D tools. My vision is sort of to build a few reusable tools that can be easily embedded in my game engine editors etc and various applications. But to do that well I wanted to do some research first. Hence, this

[NNois]

...there are more than 20 years of professional experience inside our beauty

Wrong ! Remember , softimage is 30 year old now !

[woodengun]

I say bravo for the effort BitPuffin. I had dreams of making my own SI when Autodesk made that crappy announcement but my programming skills are far from that level.

I have been using Maya, 3dsMax, and Modo for some time now, trying to replace my love for XSI, here's some of the things that have been really hard to replace:

Softimage muscle memory: Something about the keyboard shortcuts, component selection, and flow of modeling in SI has a magical combination that, any experienced user will tell you, is really hard to replace with any other 3d package. Any other program I use takes me a second to go to component mode, select, and transform; but in SI I don't even really have to think about it. (also being able to temporarily hold down V,C,X for transform gizmo interaction is a must for me!).

Simplicity vs Complexity: The original developers really made sure the simple tools worked in an elegant way. Instead of introducing gimicks like fancy viewports and sub-par FX shaders (has anyone ever used these for anything at all... ever!? come on AD....) they seemed to value some really solid foundations for modeling, shading, and rendering (partitions, and really cool compositing tools). If they did introduce a new development (like ICE) it was brilliantly simple, open, and powerful.

Why not start out simple? Just try making a really solid "SI-esque" modeling/UV package and go from there. I don't think a goal like that is out of reach and it would make a lot of us really happy!

good luck!

[Bullit]

Why not fork Blender? SoftBlender?

[BitPuffin]

woodengun: Thank you so much for your input! Starting simple is definitely what I have to do. In fact I personally believe that the best workflow is to have multiple small dedicated programs for each given task. That way you get a less cluttered environment. And then, to eliminate the drudgework of saving and opening files between them you could have some server program that shares the data between them in real time. Maybe even in a remote location using something like Verse. However since that's not how SI works maybe people wouldn't be so happy if that's what I made. Plus it's kind of ambitious so it's probably better to save as a future project or something. But yeah! I think starting with SI esque modeling + UV is probably exactly what will happen.

Bullit: A few reasons, kind of answered already, but the short scoop is: I want to learn how to build things like this myself, if I'm starting with something already built then what's the fun in that? . Second I have a feeling it would be messy to try to bend blender into being like softimage. The results would probably be lackluster. Though over time with enough boring refactoring work you could probably get it pretty nice. For myself I can't really bring myself to work on GPL licensed code if I can help it. I don't agree with it at all, I think it's wrong for people to try to control what is done with their work when it's out in the open. Few people agree with me on that and but that's fine .

As for progress update I haven't really been able to work on it at all haha. At the moment I'm working more than full time so I'm a bit exhausted at the end of the days. That said now that my sleep schedule is becoming more regular I should be able to arrive at work early enough so that I can dedicate some bus development (where I'm somehow the most productive!) to this project and still have worked enough hours when it's about time to go home! . I'm also taking a little time to research a few things to try to make up my mind on some certain coding style related things. Not so much CamelCase vs this_style but more technical stuff. Basically, I'm figuring out how to program in the simplest way in a programming language that is way too complex for its own good, so that I can document some guidelines for anyone who want to contribute code when there is something to contribute to. Because it's still important that I use a language that people actually know for a collaborative project like this. That's mostly what I've had energy left over for at the end of the days. But I feel like that stuff is starting to materialise pretty nicely. At some point you also just gotta start to see what works in practice.

Cheers guys!

[wireframex]

Hi SI community !!!
After some month to find a replacement of my precious I decided to come back from hell and to continue with SI...
Tried C4D STUDIO but there are too much things missing from there... I can't go to Maya or Max because of Autode$$$k
Im sure I don't need all new things actually. A lot of to do to know all possibility it still offers with all plugins like MOOTZ (THANKS ERIC !!!!), Redshifts .... And VRay (but dev is stopped actually ...).
All the best
Phil

[woodengun]

Hows this going BitPuffin? Any luck / milestones?

[BitPuffin]

Hey!

No there hasn't really been any milestones or luck. To be honest I've been mostly very busy and stressed with work / rent stuff lately. And most of the time after working all day I'm kind of fatigued with coding so I generally do non-coding related things. So there hasn't been much time + energy for me to spend on this. It is still something I want to see happen though, whether or not I am the person who ends up making it a reality!

Cheers!