Ara's Tale - short film wip


#61

A major milestone is reached !

I have the visuals done for scene 1, 15 shots in total. There may be minor tweaks along the road but nothing serious any more.

Finishing the shots for scene 1 was a testbed for me to discover the needed workflow and techniques to bring the shots to life. It wasn’t clear that I would ever be able to achieve what I had envisioned, but to my delight I am very pleased with the results.

This shows me that the short can be done (with a lot of work of course) and that gives me the motivation to keep on working. Further on I have now the techniques in place to tackle the remaining ~45 shots in an efficient way.

See here for some frames taken directly from the shots.

And here are the first 103 secs of Ara’s Tale. 3 shots in the inner-eye sequence are still from the animatic as they require set 2 to be ready, which it isn’t yet. The shots were rendered at 50% resolution (960×540) so you can easily scale the video up to see more details.

link to video

I will now slow down a bit after the heavy workload (and long nights) during the last few days. After that its time to start the modeling and texturing of set 2 and sculpting and texturing the dragon.

That means that updates will most probably be a bit less frequent for the coming weeks.

I also won’t show full shots from now on, but only stills, but I hope to get a trailer/teaser ready sometime early next year.


#62

Hi Martin

Looks amazing :wip:

Best wishes,

Mikkel


#63

Can you post a wireframe of Ara? (I’m interested because of rigging)


#64

Sure here are wires of the head and hands

The topology of the hands is not really good for animation purposes, but this is not much needed in my short anyway.


#65

The start into this year is a bit slow. Real life is hugely interfering with this project and then there is the blank page effect. All shots from the entry set are done and now its back again at the basics – environment modelling and sculpting.

But I managed to get started.

The task ahead now is/was the sculpting of the main set. I decided to use a different approach than the one I used for the other set. There I divided the canyon into several objects and sculpted/textured each individually. That gave quite some headaches in the texturing phase where I struggled to reduce the seams to a minimum.

This time I use one object for the main set canyon and ledge. I manually subdivided the regions which will be seen in closeup shots and thus did a very cheap ( and fixed) dynamic sculpting approach.

Instead of dividing the set into several objects and uv unwrap each individually, I created the division using seams and uv unwrapped these individual patches separately. Each patch will receive it own texture when texture painting. Then I will create a separate material for each of the patches and use the corresponding textures. The basic material settings are the same for all these materials.

I already did some feasibility tests and based on these I decided to try this approach. And now with my hardware upgraded, I will also try to test drive Mari on this setup and do a comparison to blender’s texture painting. With a 15 day trial license there should be plenty of time to do this.

Sculpting of the main part is now done. The sculpted mesh has ~850000 vertices and will be used directly as is. No displacement maps and the high frequency details will be done during texturing with bump maps. This amount of vertices is a bit high to work with fluidly in blender 2.49 (due to the infamous performance issue on fermi based consumer cards) but this is not an issue for the actual rendering. Tweaking of the animation will be done with a 50000 vertex mesh as proxy for the real one.

I wanted to have the sculpt of the set fully support the shots I had already done and for this I imported the camera rig and all camera actions of the various shots together with the Ara rig and her actions to the sculpting file. Sculpting is done in 2.5x whereas the animation is done in blender 2.49b. A minor annoyance is that you cannot directly append/link an action from a 2.49 file. You first have to open it in 2.5x save it and then you can append/link it.

With this all done I fine tuned the sculpting by having one viewport with the camera’s view and the an other for sculpting and sometimes sculpting in the camera view as well. This worked out quite well, especially where I had to ensure the correct topology of the terrain where Ara is interacting with it ( she should not be floating or wading through rocks )

See here for some images directly taken from inside blenders viewport to get an idea of the current progress. ( sorry for the wide images, didn’t want to rescale them )


#66

Time for an update …

The texturing process benefits now from the experience gained during the work on the entry set. Laying out the workflow and making decisions is a bit easier now.

As already mentioned in my previous post, I use a different approach for the main set regarding the layout of the objects and materials/textures. I have one huge mesh and use several materials/textures to cover its surface. With this I hoped to get better results with seams and evade the struggle of managing different objects with one seamless texture surface.

The approach works, but has its own pitfalls. Its mainly a question of dividing the mesh ( in terms of uv sets) so that enough detail is possible for closeup shots and that you do not have too big resolution jumps from one uv patch to another.

I started out using blender for the texturing work, but wanted to give Mari another try and so did the texturing so far using Mari. Well, what can I say. Mari is some seriously heavy piece of software. Heavy in its demands on hardware as well as offered features regarding textures. One easily sees from where this product originates and it handled my project just fine.

Now having to go back to blender for texturing is like going back to the dark ages. Some things can be done in blender but always with the extra effort or penalty in speed, sometimes to the point of unusablilty, especially when working with multiple highres textures painted on a highres mesh. And there are obviously some things you cannot do in blender but then blender is not a dedicated texturing software such as Mari is.

Depending on my future way and financial means, I seriously now consider buying Mari. If The Foundry only had a Mari PLE ( as they do with Nuke) but alas

Here are some facts on the texturing:

[ul]
[li]one object/mesh with 14 uv patches[/li][li]I use 2 uv sets one for rendering and one for painting[/li][li]each patch gets a diffuse and a bump map (each 4K), painting is done on the whole mesh at once[/li][li]I rebake the bump maps to normal maps using blender 2.5x making use of the much improved bump mapping und thus transfer the good results to blender 2.49 where the render will be done.[/li][li]the textures ( tga files ) amount to 1.2 GB at the moment[/li][/ul]Although I can cover a lot of shots with this setup there are a couple of shots where I will have to use an extra set. This is for shots where the set is seen in extreme closeup. For this I duplicate the original object and remove everything which is not seen. I duplicate it again and do a new uv layout taylored for the new shot demands. To keep continuity I then bake all textures from the old object to the new one. With this I get a very good blueprint for further detailing the closeup sets.

Whats finished now is the main part of the main set with two extra closeup sets already rebaked but detailing not yet done.

There is one thing which still bothers me, but for which I do not have a solution. I envisioned a much ‘sharper’, rockier terrain and tried to get this look with sculpting. But what ever I tried, when zooming in, the terrain always gets smooth and ‘polished’. There may be tricks to get really sharp edges and creases but I never got something really convincing. Even with the polycount at extreme values you tend to end up with smooth edges. If there is some expert blender sculptor, who can give me some tips on this, I would be more than grateful

I did some test render to see how the whole setup works and thus you are able to have a peek at some of these here.

As always during these test phases, the lighting setup is just a single spot and a hemi, so that you just can see something. Ara ( and the dragon) are not ( or not fully) textured for these renders.


#67

next update

If at first you don’t succeed …

try, try and try again.

Thats especially true for this project of mine.

The last main post showed my progress on texturing the main set cliffs and while the
basics were ok the whole setup didn’t fully work out.

As remarked upon by others at my wip on blenderartists, the textures were too uniform, having the effect that the sense of scale was reduced. My initial idea of applying fractal principles to the texturing was taken obviously too far.

Another point was the too smooth sculpt which didn’t work well together with the texturing.

With this lesson learned I went back and did a new iteration.

First the sculpting iteration. I tried several brushes and approaches and finally settled on a combination of the pinch, contrast and a bit of the layer brush to give the whole a bit of a sharper appearance. I didn’t succeed 100% but I managed better than the first round ( I guess ).

The approach I used has its limits though. Especially the pinch tool, while producing relatively sharps edges, compresses the mesh at these edges, which can produce problems when texture painting.

Sculpting changes the geometry and therefore I had to rebake the currently painted textures to the new geometry.

Next came the iteration on texture painting. This time no longer being able to use Mari, I had to go back to the tools at hand: blender and gimp/mypaint/digikam. This step back was immensely tedious. The turnaround times exploded and I had to readjust my workflow to accommodate for the longer waiting times. I might have a bit of a surprise waiting here, but I have to wait for confirmation before I really tell you about it.

Anyway, using the the textures at hand I approached the whole texturing with a new point of view. The last texture set was simple as I first created the a monochrome base map showing the basic texture of rock. This texture was directly used as bump map and a diffuse map was than derived from the base map by simply painting some color on it.
That resulted in the images shown previously.

This time I created the bump map from scratch, using a reference texture from rock walls and placing cracks and fissures were they logically would appear. The cracks were high contrast but the general texture was subdued and a bit blurred at well.

For the color diffuse map I applied the base texture using a screen operation on top of a pure color map on which I painted the color variations. I tried to do big scale variations for the color imitating big scale variations in geologic composition, like having a darker layer going with the general direction of the layers. With this you get a a better effect of estimating the scale ( I hope ).

Small scale variations are already in place through the screened base map, but to go further I used different masks derived from the base map (high/low contrast, inverted
) to paint over the color map. This gives consistent and yet not too busy variations on the color.

Doing this with a set of 14 4K maps, once need to be very patient.

But the result is better now, so the patience was rewarded

And here are the same shots as before but this time with the new textures. Again just the most basic lighting applied and the right side of the cliffs are not yet fully colored and detailed.


#68
                         [[img]http://www.loramel.net/blender_minutes/wp-content/uploads/2011/04/sc02.shot01.jpg[/img]](http://www.loramel.net/blender_minutes/wp-content/uploads/2011/04/sc02.shot01.jpg)

Now its finally time to make the first shots of the main set a reality. After a lot of fiddling with the scultping and texturing of the canyon rock walls I have now decided to continue work on the actual shots.

For a time it looked like I might had the incredible luck of getting a prolonged demo license of Mari, as Jack Greasley the product manager of Mari contacted me directly. In exchange of some material and helping a bit in marketing Mari I was to receive this license. Alas the dream was just that – a dream, but a very nice one. I do not know what happened, but after inquiring 3 times I think I should accept it and let it be a nice dream

There is still some sculpting/texturing work to do ( the dragon comes to mind ), let alone some major vfx work, but for motivation’s sake I had to start to produce finalized shots

There are quite some shots to do before the dragon actually appears, and thats what I want to do for the next time.

During the last weeks work on the score has started, and it showed that it would help greatly to have more footage to better balance the score. So this should help Phil as well.

From now on no animations will be shown any more in public. I will post some stills ( as the one above ) if I think it doesn’t give too much away, so it may be a bit less interesting to watch the progress on Ara’s Tale.

On a side note: It is really disturbing how much one forgets in just 4-5 months. I had to study the lighting setups of the first set and some of the compositing tricks I used to be able to start working on the first shot. Fortunately, during the hot lighting phase last year, I had the foresight to make intensive notes in textual form, which helped a lot. I hope that this doesn’t hint at the first signs of a hidden alzheimer


#69

It’s really nice to see this and your blog. Very inspirational. The cloth motion looks great.


#70

Ah, and I almost thought I am the only one here in this thread :wink:

So thanks Brettzies, I appreciate that :slight_smile:


#71

You’re not alone Martin, we’re still supporting you, don’t loose your motivation :wavey:


#72

Thanks, Stephan, I know that of course :thumbsup:

And I am far from loosing my motivation, in fact I have regained a lot of it by actually doing the new shots.

(goes down to the basement and switches on the 24 cores … )


#73

Great work Martin. I love how the project is coming together.


#74

Progress is quite good so far. Since my last post, I managed to finish 5 more shots. If I can keep this pace, 2011 might actually be a realistic goal.

But I have already opened a new box full of surpises, which holds the potential to delay the progress indefinitely: I am exploring the possibilities of the new blender smoke simulation.

And as with all the simulations, you have to test and test and test

Going through some tutorials helps of course, but in the end you have to know what to do, to achieve the desired effect for a shot. In the shots where the ledge is shown the first time, I made use of the smoke simulation to show rising mist coming from below.
As stated in my last post, there will be no new animations to show, but this time I give you a glimpse and show you frames directly from the actual shots:

and now, time to do some more lighting


#75

Looks good. Keep it up. Renders always help spur some inspiration.


#76

Thanks Brettzies !

Oh yes, for me as well as for those who follow this :slight_smile:


#77

I agree ! A thousand times ! Go on Martin, you’re doing a fantastic job ! :thumbsup:


#78
                         [[img]http://www.loramel.net/blender_minutes/wp-content/uploads/2011/05/sc02.shot09r.jpg[/img]](http://www.loramel.net/blender_minutes/wp-content/uploads/2011/05/sc02.shot09r.jpg)

The last weeks saw me busy on various production aspects for Ara’s Tale.

I finished ( meaning make it production ready ) 6 more shots and in   doing so, I had two aspects which I had to tackle, with one being quite a   time consuming beast.

One of the shots is a closeup of the  cage, where it is put down on the  ground. Naturally the ground is seen  in full detail here. Well, it [i]should[/i]  be seen in full detail  here. I use a 4K map for the upper part of the  ledge, which fits for all  shots, but this one ( and another one at the  very end ). The resolution  just isn’t high enough.

What I did, is to create a minimal ledge with a new UV layout and a   new 4K map were enough detail is there to get a satisfying shot in   FullHD. This involved some rebaking and another tour through Mari.
This is the shot which needs this level of detail ( which is a bad   example here as it is small and quite compressed, but you get the idea 

)

[[img]http://www.loramel.net/blender_minutes/wp-content/uploads/2011/05/sc02.shot08ar.jpg[/img]](http://www.loramel.net/blender_minutes/wp-content/uploads/2011/05/sc02.shot08ar.jpg)

I will do a separate post on my blog highlighting the workflow I used and will try to do my first screencast for this, so stay tuned

The other aspect was another one altogether .

I already had to do some minor moving mist in the previous shots. Those were done using blenders smoke simulation. It showed then, that some more work was needed to get it right and it already hinted at the problems to get the smoke simulation do, what the director wants and not just some lucky simulation results.

The shot which I finished now is the first one where you see the  first  time what’s actually down below, and that is clouds/mist, slowly   moving, similar to what you see when standing on a mountain above the   cloud layer.

Each and every attempt to get this done with blenders smoke simulation failed. At first not miserably but fail it did. My guess is, that the smoke simulation is not very good for slow moving, very slowly dissolving mist. Also the way the simulation produces new smoke constantly destroyed any feeling of reality.

I always got the puffs of smoke from new particles which could be seen very well and did not help to convey the type of mist I envisioned. There were parts of the simulation which worked quite well, but they only lasted too short to be used in the shots.

I then settled on a compromise and took the best simulation so far and tried to use it in an actual shot. This was were the attribute ‘miserably’ joined the fail.

The dimension of my environment is quite huge. And I very fast learned that the smoke simulation does not like these huge dimensions. Viewport visualisation gets weird, and the various parameters behave differently. There was no way to adapt the simulation parameters to the scale I use in my short. And a quick calculation and extrapolation showed that the cache size would explode almost into the TB domain and render times would go through the roof as well.

The other approach was to scale down the environment and do the shots involving mist separately and composite the mist back in into the final shot. It somehow worked ( from a technical standpoint) but failed again when seen in the shot. It was not plausible. It was not very well controllable either and for the detail I needed, cache sizes and render times were again in unacceptable regions.

With this I put the smoke simulation at rest ( for the clouds ) and partially for the flowing mist wisps too. I am of course no expert in this area and might have missed some important points, but for now that is how it is.

But I got my clouds in the end. Twofold.

For my slowly moving clouds at the bottom of the canyon I use several photos of real heavy clouds. Those are projected on planes which move very slowly. In the compositor I overlay those layers with itself and Sobel filtered copies of it, with the filter parameter animated over time to simulate the rolling nature of clouds.

With this I get quite usable raw footage to b used as animated texture. I also use derivatives of these clouds to create an animated displacement texture to displace a highly subdivided plane to add some volume to the otherwise flat clouds. This in combination with carefully chosen material settings and lighting is what I use now for the clouds in the shots.

For the flowing mist I use partly pieces of smoke simulation and real footage. With my brand new Canon 60D I took countless shots of fringe areas of clouds against blue sky. You get extremely interesting movements there. The blue sky is a perfect background for converting these shots into animated alpha masks and those can then be used in various ways further on.

Now that was too much text without any images so here we go: Some frames directly taken from the latest shots I did:

[[img]http://www.loramel.net/blender_minutes/wp-content/uploads/2011/05/sc02.shot06.jpg[/img]](http://www.loramel.net/blender_minutes/wp-content/uploads/2011/05/sc02.shot06.jpg)

And finally a new update on the time accounting:


    Time    Task 
  ---------------------------------------------- 
    5:34    01-Script 
   43:54    02-Storyboard 
   73:54    03-Concept Art 
   14:48    04-Animatics  
  140:54    05-Modeling   
   76:37    06-Rigging 
   179:17    07-Animation 
   179:15    08-Texturing 
   132:14    09-Simulation   
   20:32    10-Project Management  
  122:54    11-Research 
  	 6:29    Editing  
  168:21    Lighting/Compositing 
  ---------------------------------------------- 
  1164:43 Total

#79

Cool to see your progess. I’m cursious how you keep track of your time/task like that. Are you just doing it manually or based on time spent in a particular program?

Also, I noticed that this sub forum is down to two pages right now(was three), and the dates on threads on page 2 are not that old. At what point do they disappear? I know there can be big gaps between when people can work on their own stuff and not post in a thread like this. It would be a shame to see it “gone” all of a sudden.


#80

I am too am curious as to how you are keeping track of all this time. It is interesting from a beginners point of view to see how long is spend on certain parts of a really detailed animation and am curious to know myself. Doing it by hand would take seem to take a lot of time.

Check your display options at the bottom of the sub form / forum, mine was defaulted to show the last month. I bumped it up to Show threads from the beginning then hit the show threads button and it jumped to 199 pages.