A review of my project concept:
I am creating a short video featuring the piano, the instrument I play. It will be a largely procedural project in which the piano keys will be animated procedurally using MIDI (a file format describing musical notes) and particle effects will be generated visualizing each notes pitch and velocity (how hard it is hit). Above is my previs completed for the concept development class, featuring some rough shots and simple particles. The shots and timing are not final but it gives a good approximation of the look I’m going for. I think I’ll get some feedback from my film major friend on the cinematography.
Meanwhile I’m starting to learn how CHOPs work in Houdini, and particularly importing midi files.
Here’s a flipbook of some testing I was doing. In this line each point is a note or single piano key. If the note is ON, it moves up in the y axis. The spikes go higher up depending on the note’s velocity, AKA how loud it is. The next thing I need to figure out is applying each channel to motion in the piano rig. Each note must apply to a specific key and cause motion in that joint.
I haven’t done much with the particles over break but I will be working on them throughout the coming weeks.
Update 1/16/23
I have gotten a basic, choppy prototype working for the animation:
It took some time and some help from my classmate Kyle to get the animation working.
I import the MIDI file inside CHOPs. I had do write some code to run over the midi channels because midi notes go from 0-127, but the notes on a standard piano go from 21-108.
I then had to apply the data from these channels to the bones in the piano rig. I did this by group selecting them with a bounding box, then deleting everything but this group. The point numbers weren’t in the right order so I used a sort node to order them by x-axis.
I then had to apply the data from these channels to the bones in the piano rig. I did this by group selecting them with a bounding box, then deleting everything but this group. The point numbers weren’t in the right order so I used a sort node to order them by x-axis.
I gather these 88 points and assign them an attribute called @note_on, which is applied based on the channel data and their point number by a channel node. Right now I don’t have any real animation on the keys, they just instantaneously rotate based on the note_on attribute.
Next steps would be creating a smoother animation. One video I saw (linked earlier) includes animation that’s based on both note length and velocity to determine the speed. Right now my method doesn’t include velocity at all.
Update 1/18/2023:
No improvement on the key animation yet, but I have integrated it into my previs scene:
Obviously I haven’t updated the timing since setting up the MIDI so the timing is all wrong but still good to lay out how the keys might look in motion during the shots. I actually just noticed the notes being played are in the wrong key (1 note higher than they should be)! I’m not sure what’s up but I’ll look into what is causing that.
I met yesterday with Deborah Fowler, Houdini professor, to discuss my project. She took a look at my project file and gave me some tips for the animation and rendering. She suggested animating the translation and rotation of each key with a slight delay between them. If that doesn’t work she suggested I try to replicate a single predefined animation to be copied onto each key that plays. She also told me not to focus too much on the animation, and the lookdev and particles are more important.
Update 1/23/23
For the class-5 submission I have a fully timed animation and shots/layout to fit the MIDI file I’m using. I slightly altered the edit of the piece to something I think feels more fluid. I think this should be roughly the final timing:
There are a couple problems, mainly the keys still don’t really have an animation and the piano is still playing the piece in the wrong key (the animation is playing everything 1 note higher than the actual piece), but I’m planning to work on these next.
Getting the timing right was tricky because Houdini has laggy audio during playback. I had to base cuts on the motion of the piano keys to make sure the timing was correct.
Update 1/25/2023:
I have added a basic, smooth animation to the keys!
I think this is a big improvement on the previous animation and it looks pretty good at this point. To achieve this I used a Trigger CHOP, which allows me to copy an envelope, or curve onto all the note events from the MIDI file. This just means I am able to set the speed at which the note lowers and raises with each press. I also use a Shift CHOP to time the animations so that the key lands at the same time the note is heard.
I may tweak the animation more by trying to vary the speed of movement based on note velocity, but I am pretty happy with this so far. Next will be to focus on workshopping particles/fx looks and lookdev/lighting.
Update 1/30/2023:
I have been looking around at references and workshopping some particles to have emitting “from” the keys.
Here’s a look at some experimenting I did with the keys particles. Now they emit with a color based on the note velocity (cool means softer, warmer means harder). For this I copied the points of the key bones and moved them forward a bit. Using these allowed me to use the channel data to specify how much velocity to use on a per key basis.
Here is my dynamics network for these particular particles. It’s pretty simple; I am using all currently playing notes as the source point group, and varying the emission based on a note_vel attribute written to each point every frame. Then I am using a pop interact and a pop wind to add some very basic more interesting motion to them. I have them colliding with the piano with an animating VDB, though now that they’re floating mostly upwards this may be unnecessary/expensive.
Next I’m gonna focus on making a more interesting emission then single particles coming from a single point per note. I might end up doing an attribute transfer or something to get a more interesting result.
Update: MIDTERMS!
Here is an update for how my project is going as of midterms:
I have been experimenting with the note color and tweaking some of the camera angles. I met with my Professor Joe over Zoom to discuss more ideas, and he suggested some more camera angle changes as well as changing the key particles to look like they emit from where fingers would go rather than just the top of the key.
Update 2/13/23:
I have been working on the particle sourcing for the keys! Following Professor Joe’s suggestion I am trying a fingerprint-looking source for the particles. These move along with the keys so the particles are always coming from the key surface.
Achieving this is pretty simple: I create a slightly oblong circle, and scatter a bunch of points inside it. Then I copy this circle onto each of the points originally taken from the keyboard bones. Lastly I use the ray SOP to project these points onto the keys below so they are always ‘hugging’ the surface of the key. By copying onto those same bone points I was using before I am also able to store the state of whether the points in the fingerprints should be generating particles at any given moment.
The particle animation ends up looking like this (with some tweaks to the speed and various forces):
I think I like the general direction but I’m not sure if it distracts too much from the key motion, or if there’s too many. I will keep experimenting to see what feels right.
I do definitely like how it looks from further away though…:
Update 2/16/2023:
I have worked on the 3D model of the bandshell stage! I primarily am basing it off of the Hollywood Bowl but it is just a generic design. It’s super simple and abstract right now but I think it already looks good with the original lighting I had.
Next step is to pick some new camera angles which imagine the player to be human sized and not “invisible,” a suggestion my professor gave me to make my camera angles more interesting.
Update 2/21/2023:
Been tweaking the stage design, did some basic lighting testing with the stage as toruses rather than extruded semicircles as the professor suggested!
I want to definitely increase the visibility of the particles, so there will most likely be much more of them and they will probably emit a lot more light.
I was in contact with Dr. Fowler regarding the particle lighting. Right now they are using an emissive material in redshift rather than the more expensive method of instancing lights onto each one. The one downside here is that it flattens the look of the particles which could be an issue depending on the look I’m going for.
Final Version:
Settling on emissive lighting (as it was much, much cheaper), I finished up and began rendering my video. Here is the final version:
![Craig#gnick[UzjogohOpsvuftwo,2,5] Avatar](https://secure.gravatar.com/avatar/7fca9a7926d116f8eea056d4de55149c?s=40&r=r){kind=avatar}
![Craig#gnick[UzjogohOpsvuftwo,2,5] Avatar](https://secure.gravatar.com/avatar/deba7938c1861988e3c170895a6d0ecb?s=40&r=r){kind=avatar}
Leave a Reply