User:Monika/Halloween2010

For the love of God and all that is holy...

Message from the future: As of now (02:44, October 31, 2010 (UTC)) I'm at a friend's house looking for a place to watch the rally on the internet. I was at the rally but we missed a lot of the stuff that we knew would get to watch on the internet. I will be updating this while we watch but I also have a four hour drive to make tonight so I don't know when it'll be finished.

'Morning. This is me totally abusing following Uncyc's user-space policy to document the construction of my Halloween costume because wikimarkup is easy and convenient and I'm lazy. I do enjoy comments and feedback and stuff like that, on the talk page, of course.

Current plans for Halloween weekend

I'm going to DC on the 29th and staying for the Rally to Restore Sanity on the 30th (probably wear my costume) and the Legendary Pink Dots show later that evening (definitely will wear my costume unless the venue has problems with my face not being visible) canceled on account of murder. I expect to be in Durham and probably Raleigh on Halloween proper. If I'm in Durham, not busy, and awake at 3pm, I might attend the screening of Nosferatu with live orchestra at the Carolina Theatre, although I imagine theater seating in the costume might be a problem, not just for me but also for people stuck behind me. I'll probably go to the Nile show at Volume 11 that night.

sort by post age
zz 9/30/10 - You have no idea what I'm talking about. I need a this recording of this song. It's called "Nu Är Det Jul Igen" but trying to find it is like trying to find a specific recording of Silent Night. I've decided to start documenting the construction of my Halloween costume. I've been working on it for a few weeks now and some of the ideas and components were in progress last year. Instead of trying to catch you up on everything, I've decided to start this in medias res and you can piece everything together later. Here's the current status: Everything winch-related is in the mail. All the RC stuff was grabbed from a place called Toadz Toyz which was very helpful answering my questions before I made my purchase and politely dismissive afterward, although I'm willing to chalk that up to my questions on how to hack RC stuff going off the deep end. There are also some red hot glue gun sticks from some ebay store that are technically winch-related. There's also a small wooden dowel and a bit of 1.25" PVC pipe related to that system that I grabbed from Home Depot. HP has my tablet hostage. It's under warranty but there's some unrelated cosmetic damage that they might try to use to get me to pay for the broken charging system. Presumably I'll have it back some time next week. Then again, it took them 20 days to send me the prepaid box they said would take two days and which took less than one day after the final call that made them actually put it in the mail. I'm also picking up a craigslist gaming headset tomorrow; I'd decided first to do voice modulation, and then realized that I should probably consider using phoneme recognition instead of the (already 100% working as of last year) eigen-mouth vision system because then I won't need to worry about lighting the inside of the costume, which was going to cause problems with the eyes. (Also I won't have to have a webcam rigidly strapped to my face; microphones weigh less, which is nice.) Whether or not I do the sound processing myself or use some commercial software depends on how quickly I get my computer back and how nicely Matlab plays with the headset. It's been raining non-stop for nearly a week so I'm behind schedule in foam casting. It cleared up earlier this evening so hopefully I can get some more done this weekend. I need to decide asap if the gallon of Kryolan Cold Foam I have is enough or if I need to order more. This was one of the reasons this costume didn't happen last year - the company I'd ordered it from was out of stock until November and didn't bother telling me. They were okay trying to charge me for expedited shipping though, because they planned to ship it as expensively as possible once stock came in. The frame is coming along nicely. (Home Depot rocks. Lowe's sucks balls.) I'd say more, but I have a few crappy cellphone pictures I might upload today or tomorrow. I'm also considering buying a proper camera.
zy 9/30/10 - What the frame looked like before all the work I did yesterday. This is what I'll be strapped into. It's actually reasonably comfortable and I can even sit down in it (unless the chair has arms or something). The positioning of the straps makes me look like Christina Hendricks on a bad day, so you're not going to get any pictures. This thing is made almost entirely out of pieces of 1/2 inch PVC, bits of metal from the door hardware aisle of Home Depot, and white Duck-brand duct tape. The cardboard thing is temporary and used to figure out the placement of various holes in the costume front. It'll be (in this narrative - in reality I did it earlier today) replaced by the wood panel whose corner is visible in the right of the image in front of the tripod and balance ball. I made notes directly on the temporary cardboard thing. You can't read them because I used my cell phone to take this picture. The black stuff is the stuff used to illegally tint windows. For the final version, it won't be taped directly onto the door but rather mounted on a removable frame, of which there will be several to choose from to accommodate different lighting conditions and necessity of sight. The 1.25 inch PVC pipe on the back isn't structural - it's part of the winch system that I haven't told you anything about yet.
zx 10/1/10 - Most comfortable headphones ever my ass. I picked up one of these today, new, for $40. Thank you Craigslist. The reviews for this headset are generally positive, and a lot of people seem to think that it is the most comfortable headphone ever. Blasphemy. In fact, the Sound Blaster is rather uncomfortable, and I refuse to believe that that's just because I'm used to the k701; the ear-pad things are entirely the wrong kind of soft (like a mattress made entirely out of cotton balls and those air-filled bags that replaced packing peanuts) and they and the padded strap across the top are made of this crappy vinylish fake leather stuff that catches hair. But I'm completely happy with it as long as no one tries to tell me it's comfortable. Anyway, with my new mic'd headset, I've been playing around with some voice modulation software. I'm liking MorphVOX the most, but it lacks a key feature - it won't analyze a voice for you and turn it into a new voice profile that you can change your voice into. So you're left to poke around in the dark with pitch and timbre and such. Another piece of software that lacks that feature is AV Voice Changer Diamond; the difference there is that AV tells you it has that function (lies) and asks for $100 instead of $40. That was annoying. The headset also comes with voice-changing software but only lets you use its crappy presets. Anyway, if I can figure out how to get the effect I want by the end of the trial period, I'll actually buy MorphVOX. It doesn't seem fair to pirate it when I'm spending hundreds of dollars already on this project. I picked up some pulleys at Home Depot today. (I think I've been to HD nine times in the past six days.) I've begun fooling around with them but since the winch isn't in yet, I can't do much. Firefox crapped out on me while I was writing this. Surprised to find the text field survived the recovery. I'm just going to save this now and go away.
zw 10/2/10 - Remember K'NEX? A few things happened today. It didn't rain. The winch came in the mail. I didn't go to Home Depot even once. Since it didn't rain, I got the chance to work on the foam latex casting. Now, the foam latex thing is a bit of a pain - it requires a bit of experimentation and never comes out just perfectly. The biggest problem lately has been an incomplete curing at a specific part of the mold. There are a few possible explanations. It could be that the mixture isn't quite right, that there's not the right amount of release agent, that the formula isn't being mixed consistently, or that it's too cold or too humid or something like that out. Anyway, I decided today to do two coats of release agent and let both completely dry before doing the casting. It's a shot in the dark; I really have no idea what I'm doing. While it rained for a week straight, I left most of the stuff outside. I wasn't expecting the release agent canister thing to rust. Silly me. The instructions for this system tell you to measure very precise ratios of chemicals, pointing out that even a single drop can mess things up. Then they tell you to pour the thick, viscous chemicals into plastic cups... Brilliant. No pictures of the actual casting process since it all goes very quickly. Here is the aftermath - two plastic flower pots being held together with 9 lbs of weight, while a 3-dollar Goodwill mixer dries off. I asked the guy at Home Depot the other day "Is there any compelling reason I shouldn't put bits for a rotary tool in a power drill? I promise to wear safety goggles." He told me that the only major difference is that a rotary tool runs at a much higher rpm, but there's one other difference - I don't have a rotary tool, but I have a drill. Tests on a piece of wood similar to the one I want to cut show that the blade needs to run at a higher rpm if I want to get anywhere. Dremel, you win. Meanwhile, indoors... A box! Is it a camera? A real camera that takes good pictures and doesn't wake me up at odd hours of the afternoon? No, it's just some accessories to go with my 1/10th scale RC Crawler with 3-channel transmitter that I don't have. At this point my cell phone died and I didn't care, so there are no pictures. I played around with the winch for a while. The above image is taken from the Home Depot website - it is the smallest pulley they have, and it is too big. My pulley is too big. My pulley... is too big! More importantly, it is rather heavy, too heavy for the winch, and only comes in single pulley form, which restricts its ability to provide mechanical advantage. Anyway, it's sometimes useful to live in a house that the rest of my family uses as storage. Look what I found. Remember K'NEX? They were like Tinker Toys but plastic and could make your fingers bleed faster than Lego ever did. This here increases the speed of the entire setup to 500% without adding much additional weight, which is nice. Also, there's this. It came out mostly okay, but there's the same skinning problem there's been in the past, in the same place, but less severe. Also, it ended up thicker and shorter on the sides than the last one I made at this same mixture, 70g of the brown stuff and 130g of the white stuff. Might have taken too long to put the weights on or something. It's not a problem - there is room in this costume for all sorts of foam humps. If anyone lives near Durham, NC and wants chinchillas, talk to me. I have six and only want three. The other three are angsty rebellious teens who want to move out. You can have one or two of them but not all three because you don't want to make incest babies.
zv 10/4/10 - in which I ramble on about stuff and then about unrelated stuff. It started raining again today. Tomorrow, I'm going to bring in the latex stuff and just work on it indoors. I redid the pulley system; shortened the PVC pipe in the back a bit, firmed up and suspended the K'NEX thingie so it's mostly under the top of the frame, and added in another pulley loop so now there is a 7x improvement. I also realized something - I don't need the winch. I can do it manually. I can save $70! You see, when I first came up with the winch idea, I was planning to use my own arms as the costume arms, meaning I'd need something to control the thing the winch is controlling, preferably with the touch of a button I could hide in my gloves or bump with my head or control with the computer or something. The concept has since changed - my arms will be inside the costume controlling, among other things, the more cartoony arms sticking out of my head. With the 7x mechanical advantage, the pulling end only needs to move a few inches - this could easily be accomplished by a short piece of rope and some hooky system to hold it at extended and retracted positions. I won't need to worry about or lug around the heavy battery, or worry about mechanical failure or anything. But fuck it, the winch is awesome. It stays. I have however added the manual system as a backup. In unrelated to Halloween news (code for boring rant ahead), I've been sick and homebound for much of the past few weeks, had let my normally self-maintained eyebrows go to seed, and have something almost resembling a job interview tomorrow. So I decided to go get them threaded. I decide to try this new place I've seen advertised that looked like it knew what it was doing. Though a bit wary going in, I felt at ease once I saw a chart that had along its x-axis a scale of ridiculous arches (1) to nice smooth curves (10), and along its negative y-axis, a scale from Chula with a fine-tipped sharpie (1) to Joan Crawford in Johnny Guitar (10). That was good because normally the problem with these places are stocked with people who don't understand "smooth, relatively flat, and not too thin. In fact, very thick by the standards of all the creepy pictures of eyebrows on your walls." Here I could just point to what I wanted (10,7) and have it done correctly. Just to be sure though, after pointing to it, having her point to it after I stopped pointing to it, I repeated that I wanted it smooth and not too thin. In fact, she looked at my eyes and said "You already have that shape!" to which I replied "Exactly. Keep it that shape. Just clean it up a bit." And she repeated that back to me. 10 minutes later... "That's thinner than I wanted..." "I had to make it thin to get the arch in. It was flat and smooth before." Never going there again.
zu 10/5/10 - Posting from my tablet. Thanks HP! They even fixed some but not all of the case damage that I was worried they'd use to screw me.
zt 10/7/10 - Picture dump The microphone on the headphones was too wobbly but a small piece of wire and electrical tape fixed that perfectly. I haven't bothered doing any tests to see if the foam makes anything better. I'm pretty sure I'm going to do all the computer stuff in Matlab myself instead of using professional software, because it gives me more control and I'll only need to have one thing running. For the phoneme recognition problem, I've created this nice interface to collect data. It prompts me to pronounce words that it knows the pronunciation of, guesses and then allows me to refine its guesses of where the phoneme boundaries lie, and then save them for later when I figure out what the hell I need to do to process phonemes. There are a bunch of different feature vector types people seem to use and I have no idea which ones don't suck yet. This is what 9.2oz of red hot glue sticks looks like. I'm at this time torn between making the blood stream (I guess I couldn't hide that from you forever especially since it was all sorts of obvious) out of this or using this to construct the mold and then cast it out of foam. It'll largely depend on how supple the glue ends up and if there's enough to do what I want. I'm beginning to think I'm going to need more foam. Speaking of foam, I picked up some bowls today. They are smaller than the flower pots I've been using but are smoother inside. I'm also checking out advertisements on Craigslist for globes in the hopes I find one roughly 13" in diameter and hollow that I can chop up. Also, I came to the conclusion I was hoping that I wouldn't come to about the foam curing problem - I can't use the same container to mix and cast the foam. This is annoying because it means more steps and more things to clean up and more foam wasted. But on the other hand, I'm now getting good casts. And now for some old pictures I took earlier and haven't posted yet - first up, the arm-making process. Vinyl tubes don't learn their shape as quickly as polyurethane tubes, which I use in Halloween costumes all the time, do. But it's the right color and texture. I might stuff it with something, like leftover polyurethane tubing, to make it stiffer. The arms are going to be on levers that I can wiggle. This was a prototype. This project is making me glad I paid attention in 3rd grade physics. Here's an old picture of the new K'NEX pulley system. I have nothing new to say about it. Ditto.
zs 10/10/10 - Craigslist continues to be useful. I picked up a dress form today. It's in need of some repair (as it was last a high-quality dress form in the 50s), which I look forward to doing. I've actually been planning to make a custom duct-tape dress form (for when the $300 dress forms you can buy in fabric stores aren't adjustable enough and $10 isn't out of your price range) but this was cheap enough and in good enough quality that using it as a base for the custom dress form will greatly improve the quality of the final product. And it has a really nice stand (that also needs a bit of clean-up). It has a bit of cracked cardboard that would, were I not planning to cover it in stuffing and duct tape, require fixing by mounting something hard inside, and it would also need a new jersey-knit shell. But all of that can wait - for now, I've fixed it by duct-taping a piece of foamcore to the front to flatten out the cracked bit and pinning over the shell an old damaged shirt I don't wear anymore. I've put the frame on the dress form. It's the first time I've gotten a good sense of how tall it is; simply wearing the thing and looking at myself in a mirror didn't give me the full picture. I also added a plastic head, a spare from a retail hat display I happen to own. I have a Styrofoam head somewhere that I've used to do wig-work for past Halloweens that I might break out and use instead. I also picked up an old globe that apparently came with an Encyclopedia set (of books, kiddies). It's kind of cool and I might try to not ruin it and reassemble it when I'm done, but it already had a bit of damage before I got it. It's got the USSR on it and everything. It's 12" in diameter, which is shy of the 13" I was hoping for, but it'll do, although now I'm considering shortening the costume frame from front to back by .25 inches, but that's pushing it, 'it' being my ability to get inside the costume. It would, however, make the 12" completely adequate and would decrease the extent of the costume in nice ways. Anyway, the globe has been covered in liquid latex and when that's dry, I'll coat it in plaster to make a new better mold for the foam bumps. Speaking of all of that stuff I was just speaking of, I've made a mold for the blood stream out of plaster, and'll test it out tomorrow or so when I'm sure it's dry. I made a model blood spill out of clay I made out of flour, salt, and water, and the model stream is just some more PVC pipe. I covered the entire thing in liquid latex and coated it in plaster just as I plan to do with the globe, and I've broken the mold up into pieces with slices horizontal to the floor. I figure I'll start by filling from the bottom and working my way up. I have some red acrylic paint. Professionals suggest mixing it with something called Pros-Aid - I've not yet decided if I think that's overkill or not. I'll have to experiment. I haven't put much thought into painting the foam because I planned to cover it all in fabric originally, but I like the idea for the blood. And what is Halloween for if not trying new things and learning new skills? The dress form wearing the frame. I added the head for my own reference but the neck is too long. The neck... is too long! The Northern Hemisphere pre- (above) and post- (below) a layer of liquid latex. It will eventually dry to a yellowish clear. The pre-baked clay model of the pool of blood. Post-baking, with a PVC pipe stuck in it and plaster, contained in foamcore and more duct tape, encasing the thing. I honestly have no idea if this is going to work.
zr 10/13/10 - Foam Failures and Successes So I never updated you on how that plaster mold thing from the last post went. It didn't go all that well. The stem part didn't work at all, and mostly crumbled when I tried to get it off the pipe. The bottom part came out better. Figure 1 shows the result. Unfortunately, the amount of foam used was slightly less than the amount needed to fill out the mold, but it wouldn't be difficult to do it again. But there was something else wrong with it, something I really can't describe. I just didn't like it. So I won't be using that exact same method again. (I did consider using a variant of that method involving painting the inside of the mold with nail polish to get a different finish on the foam, and that's still on the stack.) Also, on that note, while I was at work, my mom's retarded dog that I'm sitting decided to try eating the original clay mold (reminder: flour, salt, water), took a single bite, and proceeded to throw up, get a drink of water, and repeat for a few hours. So if I were to redo the mold, I'd have to do it from scratch. Anyway, I also came up with the idea of letting the foam simply form the spill itself, which leads to a very nice skin different from the skin that forms with a mold. However, the foam expands and hardens very quickly as soon as mixing starts, and I'd have to pour it quickly. Now, I want kind of a cartoony looking spill shape, with kind of a bulbous silhouette, which the foam would never naturally do. Figure 2 shows my planned solution to this - a border of tape that hopefully the foam would fill in. Now, I knew that the foam wouldn't easily just take this shape, but I wanted a natural top finish, so my plan was to use my hair dryer to encourage the foam to take that shape. Figure 3 shows what a miserable failure this was. I'm putting the blood spill project aside for now. But I'll mention I'm also considering not using foam for it at all but trying out caulk mixed with latex paint instead. I'll pick up a tube for testing the next time I'm at Home Depot. On the foam hemisphere front, the plaster mold didn't work at all. I ended up using the inside of the globe as the negative mold, making it a millimeter smaller than it would have been had the plaster worked. For the inside of the mold, I found the blue bowl I was using as the outside last time fit neatly. However, I was finding that even 70/130 wasn't enough foam to fill it out a lot - the results (Figure 4) were thick and didn't cover very much of the sphere surface. I tried fixing it by thickening the bowl (Figure 5) but this only made the problem worse, since the thickened parts blocked foam expansion. I eventually gave up and went to using the other half of the globe as the inner mold (Figure 6). I expected this to be a bigger failure, but it actually turned out quite nicely (Figure 7), full to nearly the full hemisphere, and thin but not too thin throughout, even at 56/104. Also, you might notice I'm not using my cell phone camera anymore. I picked up a used Coolpix for over 90% off retail; the camera can't read SD cards so I can only take a few dozen photos at a time before dumping it to computer, but that's fine with me since I'm only really going to use it to document costume construction and sell stuff on Craigslist and Ebay. If I'm going to spend more than $50 on a digital camera, it's going to be a DSLR, and then I'm not going to be able to find a decent one for under $600, so this is what happened. Figure 1: While not a horrible failure, I didn't like the results using the plaster mold. Figure 2: My brilliant plan to make the foam spill look like a cartoon blood spill. Figure 3: A horrible failure. Figure 4: The first round of bumps made with the globe were thick and not expansive enough. Figure 5: Thickening the inner mold didn't help. Figure 6: Nesting the Northern and Southern hemispheres. Figure 7: I wish I had tried this earlier. Also, I keep name-dropping Home Depot, Craigslist, Duck-brand duct tape, and other things you all surely already know exist. The Scrap Exchange is also awesome and if you ever happen to be in Durham and working on some crazy Halloween Costume project, you should check it out.
zq 10/18/10 - Zip ties are awesome. So it's nearly 3am and I'm in bed coding Halloween Phoneme Recognition software and my cat decides to sit between me and my laptop, which is on my lap. This makes coding hard. I try to move him and he responds by doing the affectionate kneading thing, which is really just an excuse to dig his claws into my legs. Eh. So I'll post an update and go to bed. On the coding front, I am still under the impression that the best thing to do is to do it all myself in Matlab. Of course, this means I need to learn how to do a lot of complicated stuff I don't already happen to know how to do, but on the other hand, I have complete control and I'm working in an easy-to-use and familar environment. So let's see... I'm using the Data Acquisition Toolbox to deal with inputting and outputting sound. Luckily, it actually lets me send two different streams to my headset and a different speaker, and also can take in input from multiple mics. I have it currently working so that it records what I say, processes it, and plays it back with a set lag time, modulated to an external speaker, and in raw form to my headset. The "processes it" bit is still in progress but it works fine. I plan to add another mic to capture outside noise (since I can barely hear in the getup) and play that to me instead of my own voice. Output is set to lag for half a second, but I can modify that easily, and it seems to have no problems down to as short as a .03 second lag. I'm willing to extend it to as much as 1.5 second lag to deal with processing overhead but I'd rather optimize the hell out of processing as much as possible instead. Unless I can figure out how to treat video streams the same way the Data Acquisition Toolbox treats audio streams, I'm going to also have to use the Parallel Computing Toolbox in order to sync the lip movements to the audio after they've been determined. In my experience, the Parallel Computing Toolbox is a whiny bitch, but it should work. On the processing end, I have to do two things, modulate my voice and identify phonemes. For modulation, I need to do pitch modification and spectral envelope filtering. My pitch modification code is based on someone else's pitch modification code with some variations. First, I've figured out how to replace this simple "transpose by some number of half-steps" bit to instead produce a sing-songy output based on a target pitch-offset function instead of single value. That sing-songy stuff will probably be tied in to the phoneme recognition results, so I can speak in my normal, rather monotone voice and have it come out all cartoony and sing-songy. (I'm going to use that word a lot because there aren't any non-silly words for what I mean.) It's sort of like Auto-Tune with the parameters set to "producer trying to fix a bad singer's recording" rather than "Cher" (for those of you who are 12, "T-Pain"). Second, I've optimized it a fair bit, getting the processing time of a 3-second clip from 3.8 seconds (bad!) to 1.4 seconds (acceptable), largely by rewriting the phase function that comes with the Signal Processing toolbox; it had an ugly for loop and "find" that I replaced with a repmat and logical indexing, which shaved off about 1.8 seconds, and a bunch of if statements that were not needed for this specific problem because I knew what the input always looked like, which shaved off another .2 seconds. The rest was cut down by fiddling with parameters and trading quality for speed. For spectral envelope filtering, which is according to a whole lot of scientific papers about making your voice sound funny what I need to do to make my voice sound funny timbre-wise, I just need to, well, make and apply a filter using the Signal Filtering toolbox, which I can do as soon as I figure out what my target envelope is, which'll require some analysis of the original Rejected recordings. Phoneme recognition was what I was working on when my cat so rudely and aggressively insisted on showing me affection. I'm not sure how I'm going to go about it yet, and I'm trying out a few things to see what's fastest and still gets decent results. I'm currently moding and testing this particular set of tools, which was originally designed to do word recognition. My hope is that it does as well better than it currently does if I strip it down and tell it to work with less information more quickly - as is, it isn't very good, hopefully largely because its training data is what other people sound like saying words instead of what I sound like, and its training set is rather small. If it and the handful of other candidate Matlab toolboxes don't work, I might have to read more phoneme recognition papers to figure out how competent scientists would make my costume. (Oddly, I haven't seen any Matlab phoneme recognition things, just whole word recognition things. I say "oddly" because most everything has been done in Matlab by someone somewhere.) Those of you who zoned out for that can come back now. I've also been working on the physical part of my costume. Remember how my last post ended with me finding a way to make perfect bumps? No sooner had I saved that update than I started getting crappy results with that setup. Bumps were coming out too thick and too short. Seriously, WTF. But after a few of those, I started experimenting and found a happy new solution - 57/104! Sure, it deviates from the very precise instructions provided with the dangerous chemicals that'll just as soon kill someone as produce a safe prosthesis to be glued to someone's face, but damn it, it works perfectly. Anyway, now I have a bunch of bumps and plenty of spare dangerous chemicals with which to make more. Now I need to figure out how to actually mount them to the costume... By the way, I have absolutely no idea what I'm doing with this costume and I'm making everything up as I go along. ...so I attached the extra pipe insulation (left over from when I made the PVC pipe bars pressing against my back softer to the insides of the bump. Having now mounted the bump, I just need to figure out how to mount the pipe insulation thing. By the way, zip ties are awesome. I might need to go buy more now, which is strange because had you asked me a month ago, I would have told you that I had a lot of zip ties. Anyway, below are some pictures of zip ties. A pair of zip ties attaches a pipe insulation stem to a foam bump. Ditto, different view. This foam pipe structure is held together and mounted to the PVC frame entirely with zip ties. The angled corners are also created using only zip ties. These zip ties will be trimmed later. The entire structure is very sprongy without being flaccid, which is nice. There was a bit of a sagging issue with the frame that was easily fixed with zip ties. I put ones loosely looped at the upper back and middle front corners of the PVC frame, tied some nylon rope as tightly as I could between them, and then tightened up the zip ties more, which cut another 1.5 inches off the diagonal. The frame is now sufficiently rigid thanks to the magic of the geometric properties of triangles and zip ties.
zp 10/20/10 - I have a confession to make. I am beyond proficient in Matlab. I can do things in Matlab that people who are only good at real programming languages are surprised can be done in Matlab. If I don't know how to do something in Matlab, I can find out how to do it in Matlab really quickly. The only person I know who is better at Matlab than I am is Carlo Tomasi and that doesn't count because he uses magic. Its sister product, Simulink, can reduce me to tears in minutes. So why, you might ask, am I attempting to use Simulink for this project? Is it because I'm a masochist? No. I am but that has nothing to do with this. Is it because I'm stupid? Only in that if I weren't, I'd be running Linux on my laptop and doing this in C. Is it because, given the assumption that I can figure out how to do this in Simulink, Simulink is the ideal solution to the problem? Yes. Simulink is, in theory, the ideal solution to my problem. That doesn't mean I'm not one "it's 7AM and my mom happens to be in town and upstairs sleeping" from screaming in frustration. Of particular bang-head-on-wall note: The "Analog Input" block returns answers as a vector sample rather than a frame. This vector sample needs to be converted to a frame and then run through an unbuffer in order to make it a stream of scalar samples, which is what every other block in Simulink assumes it will be. The "Variable Integer Delay" block takes two streams of scalar samples, one of inputs and one of delay times, and delays the inputs by the delay times before outputting them. This would be exactly what I need, except that it works by repeating frames instead of outputting them at variable rates, so if I want, say, mouth #3 to be displayed for .25 seconds followed by mouth #8 synced to audio coming in at 11025 samples per second, it outputs "3" 2756 times at a rate of 11025 "3"s a second before switching to "8", rather than outputting "3" once, waiting .25 seconds, and outputting "8" once, making it completely useless since the video lookup can't keep up with that frame rate. "Variable Fractional Delay" would do the same except that the 2757th number it would spit out would be "6.75" instead of "8" because it interpolates the output values. There is absolutely no way to comment out a block in Simulink. Seriously. If you google how to do it, you get a lot of people saying "Oh, just put the blocks in question in a triggered subsystem that is never triggered!" only that doesn't work if the reason you want to comment them out is that they're causing pre-run (what would be called "compile-time" if Matlab compiled things) errors and you don't want to deal with them now but also don't want to lose them. I have a whole lot of nice pictures of my most recent haul from The Scrap Exchange, and what I've started doing with it, but I'm too annoyed with Simulink to post them tonight. My camera is all the way on the other side of the room. Fuck that shit. Hey, at least now you know you have something good to look forward to the next time I'm not talking nerd. Unrelated, "not food" is a really hard concept for chinchillas to grasp.
zo 10/22/10 - This post has a punchline at the end. Check out this pile of stuff I grabbed at The Scrap Exchange. It cost me $7. That's eight plastic grid thingies, two strips of elastic, four plastic stick thingies, four long flat thingies, and three things I'm told are called "embryo dairy hoops" or something like that. I have no idea what they're for, but I imagine it has something to do with cheese, likely a soft cheese similar to Brie or Camembert. But what's important is that they're made of two wooden circle things. When I bought all of this stuff, I had no idea how I planned to use it. But they all had different properties with respect to their behavior under different types of stresses. For example, the elastic is elastic, the long flat things bend easily in one direction but resist shearing and bending in the other direction and are completely inextensible, and the circular things are really really circular. Now, pictures. An early experiment - I have taped a hoop to the inside of a stem-less foam bump, and taped that to one of the plastic grid thingies that have been strapped together and draped over the costume. As a result, the bump retains its circular shape and resists forces (such as gravity) acting across it. Another view. The bump is still very soft and deformable in the other direction. In this early experiment, I attached the base of a stemmed bump to the foam grid, and the pendicel to an elastic strap itself attached to the grid thing higher up. The result is a bump that responds by jiggling tangential to the stem when touched. The following night, I start mounting foam bumps in earnest. This bump has both a hoop (zip-tied in rather than taped) and a stem. It also has a cross bar made of that flat stuff. This is to prevent the bump pivoting on the stem. It came out stiffer than I expected, but I kept it that way. It'll be good for a load-bearing lump. My camera battery was charging so there aren't many pictures of this part. After looking through all the images already uploaded onto Uncyclopedia, I have determined that this one most resembles me working on this project. This is that same bump last pictured, mounted to the frame. It is zip-tied, elasticked, and flat-thinged to the existing structure at different points. This was done using trial and error until I liked its behavior. Zooming out some, you can see a second bump on that side. The lower one, which has a hoop and a stem but no cross-bar, is zip-tied to both the upper one - the tie goes around their two hoops - and a plastic sheet at the base of its stem. The sheet is itself zip-tied to the frame. On the other side is a third bump. This one, also zip-tied to the frame, has taunt elastic on both sides and a flexible plastic stick thing taped to its stem keeping it in balance. The entire thing. The foam pipe insulation structure has been propped up on plastic stick thing struts, and is supported with flat thing suspension cables, to prevent it from collapsing under the weight of the foam bumps. (They don't really weigh a lot...) I promised you a punchline. Here it is. I'm tearing down all that stuff I just showed you. It makes the costume too tall. My costume is too big. My costume... is too big! I considered that there would be a height increase from the foam pipe frame and from the stems, but I never considered their cumulative effects. As it currently is, it makes me 6'7" tall. To put that in perspective, I have a pair of 8-inch heeled boots, and when I wear them, I am not quite six feet tall. Anyway, those past two days of experiments have given me a better idea of what results I can get with what, so I'll have a better idea of what I'm going in to with the new version.
zn 10/25/10 - The pressure is on I have, like, three days to finish this thing. Fuck. Here's my checklist: FRAME [] Make the door [] Position and cut out the eyes [] Position and cut out the mouth [] Mount Computer Screen [] Mount eye screens [] Mount door on frame [] Reverse orientation of hinges [] Get locking system in place [] Figure out placement for all internal bits [] Battery for winch [] Manual backup for winch - work out position-holding system [] Speaker and microphone [] Bag of emergency stuff and other things I'll want to carry around with me inside the costume [] Make and mount the arms. Exterior [] If there's time and foam, make small bumps for back (otherwise, use reject bumps) [] Finish all foam mounting [] Cloth cover [] Drape over and pin in place [] Mark bits that need sewing [] Sew up bits that need sewing [] Remount [] Firmly and tightly attach fabric [] Cover inside visible through eyes with black fabric Blood [x] Decide between casting or sewing - went with third option, hot glue. [x] Cast or sew [] Mount and test with winch Computer stuff [] Finish training Phoneme recognition [] Match mouths to phonemes [] Another round of optimization for speed [] Settle on parameters [] Test system for real time [] Make a backup static mouth for when it runs out of batteries or otherwise fails Misc [] Make a black cover for the headphones [] (if time) Fix / remake face mask [] (if time) Make some sort of rain coat thing that'll fit over it, just in case (forecast currently expects no rain) [] Prepare bag to keep with me (Emergency repair supplies, batteries, and such) [] Prepare bag to leave in the closest safe place (Chargers and alternators, change of clothes, and such) [] Prepare bag of stuff I'll need to set up (Costume clothes and such) I think that's everything. Doesn't look too bad. The extended slim battery for my laptop came in today, and it holds a charge well. That's good. It raises the battery life on the laptop from 2 to 5 hours. The normal battery for my laptop crapped out on me today. That's bad. It lowers the battery life from 5 to 3 hours. I might put in an order for a new one tonight if I think it'll arrive in time. I'll also need to experiment with other ways of extending the battery life, such as turning off wireless internet, lowering screen brightness, etc. I basically have all the settings to "abuse power" right now. On the bumps front, I have enough full-sized bumps to cover the sides, and have mounted all but one of them. I haven't decided on a position for the last one. I've started covering the back with small bumps cut from the reject bumps from early experiments. I really like the roundness of the ones made with the globe though so I might make some smaller ones using less foam. I've made a larger negative cast for the blood spill, but I'm considering going back to an idea I had and rejected last year - making it out of fabric. I'm absolutely satisfied with Matlab's ability to parallelize everything it needs to do. I gave up on Simulink for reasons I might explain later, and went with the Parallel Computing toolbox, which sucks. I had to trick it into doing what I want by doing 10  while costume 20  get last timestep of audio 30  parallelfor loopvar = 1:2 40  switch loopvar 50  case 1 60  Process the audio recorded 1 timestep ago 70  case 2 80  while there are frames left to animate 90  animate the next frame for the audio recorded 2 timesteps ago 100  end while 110  end switch 120  end parallelfor 130  Play processed audio from 1 timestep ago 140  Transfer information from variables in case 1 to variables in case 2 150 end while There's all sorts of trickery needed to get that thing to work. For example, every variable changed within the parallel for loop needs to be indexed by loopvar because Matlab's too stupid to realize that each piece of code (60, 80-100) in that loop only runs once, and therefore freaks out about the order in which variables might change and get accessed. For the same reason, any variable passing information from one timestep to the next must be written as two variables, with their values swapped outside of the for loop, at line 140. The Data Acquisition toolbox already asynchronizes data collection and playback, which is nice but means that any fetching and placing of data must be done outside the loop, at 20 and 130. But anyway, I've been fucking with that when I should have been working on the important part of the code, like phoneme recognition. Of course, had I been working on that, I would have been now telling you that I should have been working on the important stuff, like parallelization. You probably want pictures, and I have pictures, but I need some sleep. I'll work on getting them up tomorrow.