11 Seconds Club Thingie

  • First thing that I do with animations is to storyboard it. I'm using just plain props to get an idea of location of everything. The important thing in this part is to set the XYZ position of the toons and props, and decide the camera angles and number of takes. I don't care about poses, materials, expressions or dynamics at all at this point. I do it at 1 fps to get moving swiftly, and use oversized clothes to not get distracted to the wrong thing (or the right thing) poking through the clothes.

    Now, it's very important to set the camera locations and angles correctly, and to set the positions well too, because it's a royal pain to change those things later in the game. Also, the correct camera angles definite what you need (and don't need) to animate, so it saves time to do that right from start.

    This is my 1 fps storyboard. Whole process took only 10 minutes or so.

  • After storyboarding the camera angles and locations of everything, next step is to storyboard the main poses. I've seen people liking the idea of getting main poses on every second, but I think that it's very difficult to get natural looking continuous movements (the toons end up robotic), so I like to emulate periods of no-movement, like in anime, and therefore I'm lazy on deciding the main poses. I actually ended with just 3 main poses - one for the dog, then Terai sleeping, and Terai sitting.

    This is my step 2, still as a 1 fps storyboard. Took some 20 minutes to do:

    Step 2

    This seems to be trivial, but you need to think ahead for the transitions when doing the main poses. In particular I spent some time with that right arm, because I ended up putting her on her side, and she'll have to swiftly move to a sitting position, all the while moving the arm in such a path that will hit the objects on the right nightstand. And she'll have a blanket, so that right arm will get the clothes flying all over the place.

    My idea is to straighten the arm off-camera before the big move, to decrease the range of motion, but if that doesn't work then I'll be in trouble with the animation.

  • Hmm... trying again to embed the video for step 2:

  • Now it is step 3: to work out the start and end of transitions. The purpose is to make it clear when the character is moving and when the character is not moving. Worst thing ever is to forget to key the end of a movement, and then carry that movement over 50 frames with the toon sloooooowly moving an arm.

    Now, typically movements are in a sub-second range, so I retime the whole animation to 2 fp (it was 15 frames before, now it's 33 frames).

    So it's basically duplication of poses, and some tweaks of posing. At this point I'm looking at some 0.5 seconds precision in the poises and movement, but no better than that. Still no expressions (although sometimes I sneak an expression around so I'll not forget it later), no dynamics, no real props. It only took some 10 minutes to get through it:

    So this is step 3 at 2 fps:

  • Next step is step 4; I start to refine the movements. The idea is to keep the start and end of transitions as they are, but work out the intermediate steps to give a better movement. And to start adjust the timing to start to conform better to the speed of the movements.

    So I retime the whole thing to 4 fps. Now the animation has 66 frames, and it's still sparse on keying. When I started on Step it only had some 10 key frames or so; at 2 fps it went to some 15 key frames and now it has about 20 key frames -- and because the cameras was previously adjusted I only need to key what's visible, not invisible. So at this point it has a main series of keys for the girl, then camera switches to dog and has a series of keys to the dog, then back to the girl for another series of keys.

    I also adjusted the props on the sides to try to get an idea for their movements. The whole thing is still pretty basic, but the framework is pretty much set, and the idea is from here on to refine the framework, and avoid any revolution -- provided that right arm will cooperate, that is.

    So this is step 4 at 4 fps; it took about 25 minutes to get through it:

  • So, that's where I'm at at this moment. Next step I usually bump the frame rate to 12 fps. I used to do two passes, one at 8 fps and then to 16 fps, but I didn't really see much need to redo movements at 16 fps at all, and with time I came to like 12 fps as a compromise. So the idea is to retime it to 12 fps, and end up with some 66 x 3 = 198 frames.

    Notice that up so far I haven't worried at all about expressions, talking, blinking or hand positions. These will come at the very end. I haven't worried with illumination or any materials tuning, so that's for later too.

    But I do worry about the clothes; I plan to have two dynamic clothes - one as the nightgown, and one for a blanket. These can easily become a royal pain in animation, and I wish I could tackle them earlier, but dynamic clothes are unmanageable in Poser at anything less than 12 fps, and they are barely manageable at 12 fps. So I better tackle the clothes before I refine movements more - I may need to do some major adjustments based on limitations of dynamic clothing.

  • This is the first attempt at step 5, that is, to get the dynamic clothes under control at 12 fps. Animation has 198 frames, and took 1.5-2 hours to get the blanket and nightgown to behave. Notice the defect in the right sleeve (that needs fixing), and that in overall right side of the clothes look discombobulated, due to the wild right arm swing. I didn't make my mind, but I may choose to leave the right side as is.

    Notice I applied that idea of straightening up the right arm before the swing; that worked ok, and it's not noticeable in the animation, I think (I'm always doing that kind of cloth tweaking off-camera to get through movies in a single take).

    Notice some incidental expressions. It's easy to get carried away and start adding too many expressions and incidental movements too early - that's a mistake, and you'll get the timeline polluted. Cant add icing to the cake before the dough is cooked, so gotta focus in the fundamentals.

    Notice the clothes are not fully set yet at the end of the animation. That's fine; I gave them only 10 frames to set from default position to sleeping position (that is, video actually has 188 frames, the first 10 are for setup). That's not a worry, it's just a matter of giving more frames to setup on the initial position.

    Draft of Step 5, 12 fps, 198 frames:

  • Nice work so far mate! Can't wait to see the final version.
    This month's sound clip is very tricky, still contemplating what to do with it.
    Interesting workflow you got going ;]

  • Alright, 2-2.5 hours more got these clothes under control. I had to do a long sweep on the right arm off-camera, and some significant changes in the trajectory of the left arm, but now the nightgown stays around the arms (I think). At the end there's an incident with the right arm catching the blanket, but a minor adjustment will fix that.

    So I've spent 3.5-4.5 hours getting these clothes under control (and that means, just keeping them around the body). Most of that time was waiting for the 2 simulations to run - when they take 3-4 minutes, and then you need to repeat it 20 times, that really adds up. This is one of the things that is very frustrating with Poser - when you get stuck with something, you spend a lot of time. And I still wish that Poser would bake a simulation, and only redo from a point it changes - or that it would do it in the background.

    Anyway, I'm through the basic setup of the dynamic clothes, and this is final step 5: 12 fps, 198 frames (minus 20 frames for cloth setup):

    Now, next step is the joy of animation: refine, refine and refine. That's where you try to give some life to the characters. You work several hours with the boring stuff just to get to the icing of the cake, and I do enjoy this part. My rule is that from this point on I to avoid large changes in the framework or main movement as if I was avoiding the Devil. Rework from here on is very inefficient, and most of the times when I abandoned a simulation it was because I broke that rule (and I break it all the time, hard to resist).

    I plan to get the final props and environment setup last, as I find those very boring.

  • By the way, critics are welcome.

  • Maybe offset the audio a few frames later, when the stuff starts flying.

  • @krios said in 11 Seconds Club Thingie:

    Maybe offset the audio a few frames later, when the stuff starts flying.

    It's funny you referred to that. Suddenly my audio started to get out of sync with the movie. The audio generated movie is now coming a full second behind the audio in preview. I have no idea what suddenly caused that. After several attempts I just left the audio in this last movie at a random start, and let it to hopefully fix that later.

    If I can't get the audio sync'ed again I'll have to abandon this project, as it depends on exact timing around the sound.

  • @fbs7 Looks good so far.

  • This is the timeline at this point, that is, camera takes set, main poses set, main movements set, dynamics set:


    Notice a few things that tend to repeat to me quite a bit in animations: Poser doesn't allow one to annotate the timeline, so I need to keep the following things in my head (which ends up being the limiting factor in compexity):

    • I use one particular camera as the movie camera (is the aux camera, as the real movie camera is set up as a dolly, and I don't like it), and I keep track of the takes in the timeline by keying that camera. I tend to think in two kinds of takes: one that starts/ends movements, and one that is incidental to show a different shot of the same movement. So the blue arrows are start movement takes, while the red stars are continuous movements that need to have continuity in the camera.

    • Notice as the main camera takes start, the focus of the movements change - red blocks are for Terai and her hair, and blue blocks for the dog. I see that happening all the time in my movies, so one can keep a bit of sanity by identifying those blocks in the timeline and try to remember who is doing what by looking at them, but it would be so much easier if we could annotate the timeline.

    • Notice a few keys in toons that are not in the main movement blocks. These are off-camera setup moves, to get the toon ready for the next take before they start. Usually I do that on the frame immediately before the new take starts, but with dynamic clothes you need instead to spread that out. I try to keep these as a minimum, as they tend to get confusing fast.

    So that's how it works for me - I tend to switch to/from timeline window, and identify where I am in the timeline window by noticing those patterns. That's why I like 12 fps - the timeline is much shorter, and it's easier to view it all.

    This is still a very simple animation, and you can see that from how sparse the timeline is. But now it comes the time of adding the "little movements".. and that's where the trouble adds up very fast, because the timeline starts to get polluted with all sorts of incidentals, and it's very easy to lose track of what one's doing.

  • Oh, and I forgot to note: the yellow block is for the first set of stuff tumbling down, and the violet block is one last thingie that I added falling late to answer for a late "thump" heard in the soundtrack.

  • Check this action out:


    This is the 2 FPS stage, for a 2 min. sequence. Keyframe organization is an absolute must at this point, or down the line (24 FPS) things can become very messy. Some kind of annotation would be priceless, but until that happens, humans have pattern recognition: notice those stars through out the timeline, they are markers, and have no useful keyframes. You can easily tell the characters (green) from the props (gray), and also where each shot begins (spline break), and ends (constant key).

    And as you might have guessed, the last row is for the camera, Dolly in this case. Also helps denote each shot.

  • @krios said in 11 Seconds Club Thingie:

    This is the 2 FPS stage, for a 2 min. sequence. Keyframe organization is an absolute must at this point, or down the line (24 FPS) things can become very messy. Some kind of annotation would be priceless, but until that happens, humans have pattern recognition: notice those stars through out the timeline, they are markers, and have no useful keyframes. You can easily tell the characters (green) from the props (gray), and also where each shot begins (spline break), and ends (constant key).

    That's a marvelous idea! I had never thought of that!

  • @fbs7 said in 11 Seconds Club Thingie:

    That's a marvelous idea! I had never thought of that!

    If you keep your keys organized in this way, when you go to 4, 6 or 8 FPS, it will all look the same, except more keyframes.
    In other words, don't leave any blank frames until you are ready for final frame rate 12/24 fps.

    If you're interested, I'll post the progress of this scene for you... next step being 6 FPS.

  • This is the first draft of step 6 - give incidental movements. My keyword here is subtle and make no harm; of course I couldn't follow that, I added some leg flailing, which sent the blanket flying, so I had to spend 2 hours fix that (I swear I hate Poser's cloth simulation, I'm sure that would be a 10 minutes task if I could simulate the clothes in Marvelous Designer).

    Anyway, I ended up doing the old switcharoo with the blanket: there are two blankets now, one for the initial static position, and another that gets visible when she jumps up. That 2nd blanket got the sides morphed with the morph tool to positions on the side of the bed, and then these sides were made chroreographed; that anchors the blanket to the sides, and stops it from flying out (that's called pinning in Marvelous Designer).

    I also found the reason why sound was getting out of sync - Poser bug, of course. If you set animation start at frame > 1 and set sound start and another frame > 1 then the movie will be generated with sound out of sync; so I had to get the 1st blanket simulated with the initial sleeping position to frame 20, then create morph called FitAtFrame20, then set that on frame 1 and get rid of the fitting frames as these were throwing the sound out of sync.

    So I should be working on incidental movements, but instead I worked only 15 minutes on that and spent most of my time fixing sound and cloth simulation. Anyway, for the incidental movements the idea is to have the head move ahead of the body (eyes always move ahead of the head, and head always move ahead of the body... if you move the head with the body that gives a robot motion), then small variations around the paused positions (or the toon will look like a statue). So I'm halfway done with that. But all has to be subtle; the head prep move for example is just 2-3 frames ahead, and the moves to unpause the body are just lke 2 or 3 degree rotations. I also added a small jump as I thought that would be fun.

    Notice there are still many small artifacts with the clothes - intersections, parts get stuck with each other, spurious collisions, etc... I'm leaving these for now as I think I can fix them at the end when I'm done with all movements, and maybe a better simulation will help with that. I might even retime to 24 fps for the cloth dynamics alone.

  • @fbs7 Not really a suggestion, more of an alternative submitted for consideration.

    Here's how I do cloth sims:

    Firstly I'd hide cloth items & finalize the animation, then I'd be re-time to 30, or sometimes for quick movements, 60 fps for the cloth sims(you can re-time back to 15fps later if you like, I prefer 30 personally), but I wouldn't use that simulation...after simulation runs create keys every 30(or 45 or whatever) frames (match the keys of figure if you like, but not necessary), then spawn a morph target(name morph frame#_, don't forget frame1) at those keys (you can spawn 'tweeners' & edit any of those with the morph brush if needed, use a minimalist approach only fixing poke-thru) then clear the simulation, or better yet export the item and import a new object and copy morphs from original(hide original).. Double-click on each morph, set min0 max1 (force limits if available) set value at 1 for that frame and check that keys before and after are 0. Sounds like a lot of work but... viola, near perfect (smooth & editable) cloth simulation.

    When it comes to re-timing your animation , parent your cloth and/or other items to your dolls 'body' (never a body part, as movement of body part may be transferred to the cloth or other items) any changes made in re-timing the body will automatically transfer to the rest of your scene.

    I'd post an example video, but everything I have is NSFW and may violate NDA with other parties.