Currently the animatic is set up a little weird!
We’ve linked in scenes from every blend file into a master timeline in the video sequence editor (making significant use of the mapping cameras to timeline markers)- which is both cool and not-so-cool. On the plus side, it gives immediate updates. Whenever anyone updates an animation or a model, all we have to do is update SVN and bam! It’s in the animatic.
The downside to this is that the VSE was never really designed to do this. It can, but from what I hear the feature was implemented more as a joke than anything. As the files have become increasingly complex, the time to open/save the timeline has increased to rawther absurd lengths, and since it has to load each scene in order to play it in the timeline, there are often long lags between cuts, which makes judging timing difficult. I have to render the entire film in order to see how the timing actually works out, and it shouldn’t surprise anyone that that can take a while (Although, huge thanks to Campbell for putting in an ‘openGL animation render’ option in the VSE, like we have in the 3d view! Without it, this wouldn’t be feasible at all).
All that to say, we’re probably going to switch over to the Sintel method soon, which consists of rendering AVIs out of each individual scene, and just editing those AVIs in the master timeline.
Over the past couple weeks most of my time in the studio has been composed of making the animatic as accurate a representation as possible of what we’ve got to capture on set in 3 weeks. It’s a very iterative process. Initially all of these thoughts were just conceptual: ‘A shot of Captain sitting at his desk’. Now we have to figure out where exactly that desk is on the set, what type of camera movement we can feasibly get on set, and what type of camera rigs/lenses/lights we’ll need to film the shot.
This is also the part where we get to look at the film and make sure every shot’s composition is as interesting as it can be. Of course, we’ll have other chances to do this as well (again on set, again during the edit, and again as we set the shots up in final 3d space and composite it all). I think it’s important to take advantage of those opportunities. While it means the end film may not resemble the opening animatic as much, keeping the film a bit ‘soft’ ensures we’re able to put as much creativity and good ideas in there as possible.
As a quick example of how this process goes, I figured I’d upload the latest version of a shot I put up a while back.
1.)
This was the original storyboard. Panels 23-29. Gorgeous stuff by David, and I loved the big sweeping shot, but if you’re going to have a big sweeping shot, I felt it needed some sort of justification. In this one, it started strong, with the boat, tower, city, and sunrise, but ended on just the tower. Since this is the first look we get at post-apocalyptic Amsterdam, it would almost make more sense to reveal in reverse- but since this is the establishing shot of a scene which takes place in the tower, it’d be a bit weird to fly away from it, just to snap back in.
2.)
This was the first animatic. Simplified a lot, and designed to lead the audience into the scene. 1.) establish a brooding dawn (not really visible in the animatic) 2.) introduce an unknown shape rising from below 3.) reveal it’s an augmented clock tower 4.) and then there’s a flying boat thing. The actual composition of the shot is more-or-less the ending (or, the concept of a large shape passing behind the clocktower), and everything’s working towards building that image.
3.)
This is where the animatic is as of this morning. The same as before, but we’ve added one more reveal/element. And we were able to get that big sweeping shot back in! Now, though, it’s revealing the crumbled spaceport, which should really drive home the contrast between this and the previous scene, where we see the spaceport shiny and operational and not fallen over.
Even still, there are a few things about the composition that don’t thrill me. The ending has an unclear point of focus, it’s a bit too symmetrical, and there isn’t all that great use of negative space. There are also a few animation issues: the camera motion makes the boat movement look like parallax, and the boat probably enters the shot a bit too soon to begin with. But! That’s why we keep working at it!
very impressiv I like the Skyline… the clock Tower the ship (everything ;))
who are you going to realise the short by Camera Crain or thad cool smal Helikopters or comületly in Blender????
We’re gonna be using a whole bunch of techniques. We’re trying to schedule a trip up the actual clock tower during the right lighting, to snag some panoramas, and we’ll probably combine those with layers of camera projection, with some fully modeled foreground buildings… we’ll see how it goes!
Oh dear sounds like a lot of work-fun-stress-wonder experience
Perhaps the air boat should keep moving to the right at the end of the shot. Feels a little awkward that it just stops. And if the camera continued its’ path of movement to the left it might be a cool place to wipe the screen with an object like the roof of a house or building in the foreground to the next shot. Looking really cool.
It doesnt stop.
But it definitely looks like it. It’s a great example of a problem that wouldn’t jump out at you just planning shots in your head.
ok great stuff
thats incredible feature @(the beggining of blog), from wat i understand that is able to speed up the modelling to animation pipeline dramatically i.e. so that animations don’t need to be re-animated/re-tweened again all the artist needs to do is improve on the rigged charecter/set/prop and then let it update in the video editor and bam the final animation is ready……and ready for compositing……CAN THIS BE AN ACTUALL featue in blender…….that would be awsome OR even better SHOW US HOW YOU GUYS DID IT TUTORIAL OR SOMETHING….
PSS ton should make a cameo appearance (or even all of you walking into the scenes at differant times)
As far as I know, it’s been in there for a while! You can just link a scene into your timeline file like you would any asset, and in the VSE just go “Add” then “Scene”- or something like that.
imagine if the animatic can be change with re animating the entire animation…..
are you guys using SSDs
very interesting !
I like the idea of being able to link scene shots in the edit timeline. I think for little project with not so much geometry it may work nice. And the new option that Campbell has implemented will be very useful, I already wished blender has it. Thank you for that !
What about keeping that workflow and use the proxy tool to “bake” the scene shots ? it give a 50/50 between that workflow and the one used on Sintel. So you can re-generate one specific or all shots in one go, during nite or lunch :)
and there shoudn’t have lag between cuts… that way you can see all improvements on shots without re-rendering them manually.
Also that nice to see the evolution of the clock shot, it’s getting better and better, already very impressive and that’s only opengl :)
Since this is something that will be done on a near daily basis, perhaps having the task of rendering out, and updating the animatic made automatic to some degree would save a lot of time, and hassle.
Perhaps having a small script that renders out the OpenGL view to a set folder, and then have another part of the script initiate an update (check all time stamps of files in folder, or revision numbers) when loading up the animatic .blend file would allow for seamless updating with out too much user interaction.
Slightly off-topic –
It would be nice there was an option to attach Meta-data to datablocks, images, videos etcetera, and for Blender to be able to read this.
In this scenario you could add stamp information that is visible (owner / artist, scene / shot, revision, stage…), which would be human readable, and meta-data information with the above, and a few other bits of information which would be non-visible, and contained within the file.
Using the scene / shot numbers you could then add a meta-data block into the VSE which basically looks in a specific place for certain data, once there is files in there it simply loads them up and shuffles the VSE timeline accordingly, so if the meta-data block only takes up 10 frames, but the video is 500, then it would shuffle everything after the meta-data block 500 frames.
The above is perhaps not the best example, but I do think that allowing an artist to embed information via Blender, and for Blender to be able to read, and understand that would have many great benefits. Especially if Blender were to have an asset management type system in place in the future.
1) are you seeking any community models like in Sintel?
2) How many easter eggs are scheduled to be in this movie?
regards
TFS
My thoughts regarding composition:
I’d use a slightly wider lens and place the camera slightly lower. I’d time the ship to come in after the clockface is recognizable, and fill the frame a bit more, in shadow for contrast. On its way, it may pass by some lights that just give away bits of detail of the ship’s exterior, and by the end of the shot, it should appear smaller in the distance. Finally, the clockface and the ship would be at opposite corners of the frame in a rule-of-thirds position to offset the uncomfortable symmetry. Compared to the animatic, it would be nice to have more contrast in terms of lighting.
My three cents. :)
Hrmmmm,
I prefer Clock Tower Early Previz, with the added suggestion that the camera should pan down and be farther so you can basically go from the sky above to showing Post-Apocalyptic Amsterdam.
The airship can still come in, but it would have to be farther away.
If the blocking is correct the Clocktower (which I think you are supposed to go “into” after this shot?) will dominate the shot from its spire coming into frame up to about just as its clock face is near the top of the frame.
You can then have Rest-of-Amsterdam below.. the airship can then come in to add feeling.
Just my two-cents. I already love the idea behind Mango.
Hmm, not really keen on zooming/reframing lens during a shot. But I guess everything old is new again.
I have been generating proxies from Meta strips or scenes and that works well, but seems to not work lately.
Have read that other FX houses get an artist to render different levels of finish as they go, so a scene starts with blocking and is layed into an edit. Then over time is regenerated to better quality, but the strip in the VSE will remain the same just updated over time.
Can you use multicam strip for varying types of shot finish? That is, a camera strip for low res proxy objects, another for color version, another for high res and yet more for comped effects etc.
Will your animatic consider speed of shot as well as composition, as there’s no optical flow / decent speed retiming in blender and it’s better to do it in camera anyway, it would be good to be aware in advance which shots need shooting at higher frame rates /shutter angle?
Ian, what happens if a change in the composition right at the end calls for a stretch of the shot, altering the timing of some elements? Would you break picture lock and render a bit longer? How close to delivery could you continue to do that?
Typically, audio is the big thing that defines picture lock. We can either lock the whole film, or just certain scenes, but once that happens, a change in edit requires re-timing of the audio (even worse the scene’s already been scored). Until audio, though, it’s not that big a deal.
We’re not going to have picture lock for a while, though. At least two weeks after the shoot (and very probably longer, as we’ll have to be developing the workflow at that time as well)
Would the shot work if the camera just panned down and out, without so much (any?) rotation?
It seems that the rotation is the main cause of the parallax problem.
The camera feels like it’s moving too fast. There’s a lot to be looking at for that amount of time especially if the sun will be there too.