Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Vast Scene Backgrounds + Low Memory Usage = No Prob
#1
FrameForge is previsualization software that gives you the power of using actual 3D models to create a storyboard that represents your vision. There is a lot of power in that. Other programs that use clipart or static 2D imagery simply cannot compare to the flexibility of of 3D models you can pose or texture as you need. Of course, power comes at a price sometimes and in our case, that price is the cost imposed on your computing hardware in order to render those 3D models. FrameForge is very good at rendering the scene (as close to) in real-time. When you pose an actor's arm, you immediately see it rendered & reflected on the screen. Having said that, FrameForge is software and as you know, software is fairly worthless without the hardware to run it on. Hardware requirements for FrameForge are fairly low so that almost everyone can use it with fairly modern hardware. Of course, having a lot of RAM, a powerful GPU and a modern CPU is going to allow you to very grand things with the program but if you are closer to the minimum hardware requirements, you can quickly find yourself placing a burden on your hardware that it is not able to handle.

Recently, I worked with a user who uses FrameForge on a Microsoft Surface. That's a tablet computer! Very impressive work, Microsoft. Still, while it worked well for the user when he was creating small intimate scenes, he ran into a situation where his Surface ran out of available memory while trying to create a scene with a vast, expansive background. In generic terms and so I don't give too much away, lets just say that the user was creating a scene using objects from the World Clothing pack. A scene with roughly one hundred medieval knights on horseback. As the count rose towards 100 on the set, the Surface slowed down until eventually it ran out of memory.

Now, there isn't much I can do to solve a user's "out of memory" problem. There is really no way I can add more memory to the user's computer, that is not an option. Therefore, the only recourse is to come up with an alternative that uses less system resources, i.e. memory. Using the actual model of the knight on horseback, using so many of them, is blowing the user's systems resources out the window. There are a couple of approaches which can be used which will consume considerably less system resources while still delivering the backdrop the user wants. These are "work arounds". They're cheats or hacks or whatever you want to call them but they allow you to create the scene you want. By the way. there is nothing wrong with cheats or hacks in the film making business in any case because it happens all the time. The puppet that played the role of ET the Extraterrestrial was a hack/cheat. CGI is a cheat or hack. When George Miller made Mad Max: Fury Road, he didn't hire thousands of extras to stand around as crowds, he used CGI to duplicate actual extras many times over to create the illusion of crowds.
[Image: FAqMw0I.jpg]

A matte painting is a painted representation of a landscape, set, or distant location that allows filmmakers to create the illusion of an environment that is nonexistent in real life or would otherwise be too expensive or impossible to build or visit. This is the approach we suggest you do with your scene when you want to create some vast, expansive background. Lets continue to use the scenario of a hundred knights on horseback. We will create the illusion of a hundred knights on horseback galloping into battle (or whatever) and here are a few ways to accomplish that.

Use an Image to Create a Vast Backdrop
This approach works pretty well if the objects of the scene are not moving around the set a whole lot. I've used it a lot where the scene has the actors progressing through some dialogue of some sort, like a romantic conversation for example. You can view an example of what I am talking about, plus a short tutorial, by watching this video on YouTube.
<!-- m --><a class="postlink" href="https://www.youtube.com/watch?v=g6-5VBQuZDg">https://www.youtube.com/watch?v=g6-5VBQuZDg</a><!-- m -->
In the scene you want to shoot, you would need to grab a number of random shots of knights on horseback in different poses and then use Photoshop or some other bitmap editing software to create a collage out of these shots. You'd save the final result as an image, place it into your texture library and then apply it to a cube (found in the geometric shapes section of the building blocks props category) as a texture. Of course, the cube needs to be scaled to match the dimensions of the image you create. I've created an example of this, took me about 8 minutes from start to finish so I didn't put a ton of effort into it but the result isn't all that bad.
[Image: QKkhC23.png?2]
And this scene is composed of 4 objects: A texturable cube, a knight, a horse and a virtual camera.
[Image: wzOa4U2.png]

Bear in mind, that like all things, there are pros and cons to this approach. One of the pros being its fast and easy to deploy. On the flip side, its not always that flexible. If the scene you intend to shoot requires multiple cameras and a slot of camera/angle changes (as is common in a battle scene for example) a single background image or backdrop might be too static.

Using FrameForge "Plane Objects" To Create Low-Resolution Background Objects
This approach is similar to the one I just discussed but is a lot more flexible. Included in the Props inventory is an item called a "Plane Object". It's like a transparent wall you can apply a texture to that will always face the active camera. If you place one of these objects in the middle of your set and then place cameras at all 4 sides of the blueprint view focused towards the plane object, no matter which camera you switch to, the plane object will appear to be turned to face it. So what you can do is, again, snap shots of knights on horses in various poses, export those images and use Photoshop or some other bitmap editing software to remove the background of those images leaving just the knight and horse. Once done with that, save the images a PNG files with an alpha mask (transparent) into your texture library. You can then apply these images to plane objects as needed and duplicate them as needed to create an even more crowded background. I spent about 10-12 minutes doing this and came up with the following scene.
[Image: QKkhC23.png?2]
Again, the only objects on the set are a knight, a horse a virtual camera and a bunch of plane objects I duplicated a few times in order to create the "crowd".
[Image: sag1vLL.png]

Again, pros and cons. Pros = fast deployment and a lowered burden on your system's resources. Cons = since the plane object always faces the active camera, spinning one of them to reflect a different perspective doesn't work. For this solution to work in a scene requiring multiple perspective changes, you need to create a number of poses and snap them from various angles. As an example....
[Image: EVJTkLK.png]

Using a Combination of These Techniques
There is nothing stopping you from using both techniques I've outlined above at the same time in your scenes. Doing so may provide the additional flexibility in crafting your scenes that using only one on of the approaches would prohibit. Taking that mindset even further, there is nothing to stop you from using the plane objects and some crafty Photoshop to create groups of riders represented by a single plane object. Just like the first technique I discussed used an image that had over a hundred riders, you can apply an image representing multiple riders on a plane object. Like so...one plane object, one actor, one virtual camera.
[Image: xjSSjJO.jpg]
[Image: GJ37weo.png]

And finally, if the techniques I've outlined work well in conveying the idea that there is an army of riders a fair distance away from the focus of your scene but not so well in conveying riders closer to the focus of your scene, you can always add in some of the actual 3D models of a knight on horseback to represent riders in the "near background" of the scene. What I am suggesting here is trying to strike that balance between the scene you have envisioned in your mind and the limitations of your computing hardware.

Some Key Points to Remember to Implement These Techniques
* When I snap shots of actors in various poses to use as backdrop images or with plane objects, I change the set parameters so that the sky and the ground have no texture and I change the color on both to the same bright green. This makes it so much easier to remove all of the background from the image in Photoshop so my end-result image is only the object/actor I want. It's the same concept as your local weatherman giving the forecast in front of a green screen. The green screen gets replaced with a map of the area with temperature or the radar view of clouds, et cetera.
* In the first process where I talked about using a backdrop image like a matte painting, the image I applied to the texturable cube had only riders in it with an alpha mask (transparency) over the rest of the image. This allowed for the riders to blend into the set's environment easier but there is no reason that the image could not have also represented an environment of its own. The trick is, of course, to match the environment of the set to the image on the cube (or vice versa).
* Whenever I use a texturable cube in this fashion, after I apply the texture to one of the faces of the cube, I drop the opacity for all of the other sides of it to 0. This of course makes the remainder of the object transparent. In some situations you may see a light haziness in the applied image. This is most likely the result of the sun light source reflecting off the object. Changing the cubes Shininess level to 0 in the green room often take case of this problem.
* The plane objects I spoke of can be found in the props inventory, in the Building Blocks category, under the Bill Boards subcategory.

So that's my take on the whole "giant massive backdrop" shot. I tend to look at FrameForge like a sound stage and if I can cheat or hack a solution that is going to save me time (& money) then that's the approach I am going to take. If any of you have a different approach to this type of situation, I'd love to hear about it. Just post a reply to this topic. Anyway, I hope someone found this helpful. If so, follow me on Twitter, friend me on Facebook, subscribe to me on YouTube, stalk me on Instagram, watch my Vines, endorse me on LinkedIn, cyberbully me on MySpace, review me on Yelp, join my circle on Google+, pin my Pinterest, check out my Tumblr, follow my Flikr feed, slam my political ideologies on Reddit, meet me at a Meetup, criticize my creative side on DeviantArt, praise my quirky funny sense of humor presented in animated GIF form on my Imgur feed.
[Image: c9Cuvhp.png]
Reply
#2
Thanks so much for posting this!

It is extremely clear--and enormously helpful. And the video's great.

You just enhanced both my work flow and my productivity.

I'm definitely doing this going forward.

Awesome post.

Thanks, Chris!
Andre Tittle
Reply
#3
I am glad you posted this, and I hope you do more in the future.

Nonetheless, I still advocate that the code for FF be re-written so that it can handle higher polygon counts. In other words, if Innoventive revamps the application, users should no longer have "out of memory" problems.
Reply
#4
hello

Could you convert all library in Low Polygon (specially objects for war: inlcuded World Clothing Pack and Reel Faces Pack). so that I can create a (small) battle between 2 armies.


thanks for you help
Reply
#5
Banksyde Wrote:I am glad you posted this, and I hope you do more in the future.

Nonetheless, I still advocate that the code for FF be re-written so that it can handle higher polygon counts. In other words, if Innoventive revamps the application, users should no longer have "out of memory" problems.

It's not as simple as just choosing to "re-write the code" to handle higher poly counts. That implies we aren't doing everything we can to squeeze as much as we can out of the memory available, and we already have highly efficient code that keeps getting faster and more efficient, with each iteration.

There is no magic bullet, nor is there a rewrite that would simply end memory limitations, as much as we might all wish there were.
Reply
#6
that problem disappoint me. I can't organize a battle between two armies. this is a only but major imperfection of frameforge " great Software ".
Maybee the new version fixe that problem... I hope so!
Reply
#7
I don't understand. Applications such DAZStudio and Poser seem to handle higher polygon counts and neither are near as sophisticated as FrameForge. I am not asking for a magic wand. And I don't think I am asking for anything more than most FrameForgers would ask. It would be nice if FrameForge could be a little more robust and handle higher poly counts.
Reply
#8
I cannot wait for the new version fixe that problem. I use now DAZ Studio 3D.
Reply
#9
I like DAZ Studio. It is a great resource for assets. I use Carrara to modify the assets. And then I import them into FrameForge.
Reply
#10
Ill post a comment here because Chris has helped me enormously in the past two years since I started with FF. I have the latest 3.5+.

I will say that I had pushed FF towards its limits with my Surface Pro 3, and had crashes with its Intel GPU. I abandoned that hardware and bought a top of the line Surface Book. This PC is an i7 dual core with 16 gb ram and a 1gb vram NVIDIA GPU.

Believe me, as Chris can attest, I push my scenes to their limits. I wanted a "garden in a forest" and there are 600 trees, entire high polygon suburban neighborhood with houses from SketchUp, high polygon flowers from SketchUp. These sets are probably much more massive than the battle scene here. I can tell you with my setup, it's ALL ABOUT the GPU. I am very happy with the performance of FF, as long as I don't push it to the limit with lighting/shadows on AND a dual monitor displaying a camera.

I typically run a GPU monitoring application to see just how much GPU memory I'm using. I run about 400-600mb of GPU memory with lighting/shadows on my most complex sets. That DOUBLES when I have a second monitor activated, slows down and does crash sometimes, but that's always due to running out of GPU memory it seems from my tests

Chris actually had a better suggestion for hardware--a video desktop workstation--which would have cost me less and had even more GPU memory. I suppose if I had invested in a top of line desktop with the capability to jam in a top of line NVIDIA (gaming with 2-4 gb), I could push that envelope as well. However, I needed something portable. A real gaming laptop with more memory on the GPU would have been better, but I liked the Surface Book.

I DO understand that cheating the scenes does limit the flexibility of camera angles and one has to account for that, but Chris always comes up with a good cheat that looks great. I decided to just spend money on hardware.

For people who are going to be going further and doing video work with say, Adobe Premier, or After Effects you'll want a great video graphics workstation to start with.

And again, Chris, thanks for pushing me to upgrade my hardware. I may just go a step further and buy a high end graphics workstation and then I'll be asking you for recommendations!

Alan Knittel
Writer-Producer
Phareton Media
Reply
#11
Perhaps I am naive. Nonetheless, I am curious. Why is an application such as Poser, which has a number of similarities to FrameForge, able to handle objects with higher polygon counts than FF can?
Reply


Forum Jump:


Users browsing this thread: 1 Guest(s)