weaving

July 23, 2024 by Tien Chiu

Weaving faces

I’ve spent the last week or so working on the cyborg design. It’s technically very challenging. That’s because weaving a realistic human face is not easy!

While we humans have only vague ideas of what most objects look like, we are VERY familiar with human faces. In fact, our brains are physically wired to recognize and “read” human faces. That means that, if you want to produce a convincing human face, you need to be spot on about things like skin color, facial features, and proportions.

So I’ve been doing a lot of digital swatching to come up with yarn colors.

The first step in this swatching is to come up with a starting palette. I sampled the colors in the image I’m weaving, looking for colors that would mix into the colors I wanted.

Here’s the image I’m weaving. (The head is human-looking on both left and right sides because I will be sewing on the cyborg parts later.)

Then I selected five colors from the image. In the double satin structure I’m using, I can use multiple wefts together, so I constructed a set of swatches that shows all the mixes of 1-2 weft colors, like this.

(Top and left rows show the original color palette; the other rows show combinations of those colors.)

color swatches showing different weft combinations

That gave me some ideas of what the weft colors would look like when blended (in practice the blends are not nearly so neat), but didn’t account for the black warp color. Because of the structure, black would blend with the weft colors. I could have 20%, 40%, 60%, or 80% black on the surface, with the weft color making up the rest.

To simulate this, I added a black overlay in Photoshop, with black stripes at 20%, 40%, 60%, and 80% opacity. The overlay (by itself) looked like this:

black overlay to simulate effects of structure

Placing the black overlay over the extended color palette gave four shades for every weft color combination – 20% black, 40% black, 60% black, and 80% black, for a total of 15 x 4 = 60 possible colors.

However, if I tried to use 60 colors, especially nearly-identical colors, in a palette, I’d drive myself crazy (for a variety of technical reasons that I won’t get into here). So I winnowed them down to 18 colors:

Then I told Photoshop to simplify the image to a version containing only those colors:

It’s pretty good. You can see some graininess in the photo, especially around the tip of the nose, but it’s not bad.

I wasn’t entirely satisfied, though, as I felt the face didn’t have the glow of the original. Mixing in the black had darkened and reduced the saturation of the face. So I went back and created a new palette, with colors just a trifle more saturated and lighter than the previous set.

The new palette did significantly better – the version of the photo using that palette is below – so I decided to roll with it.

photo flattened to second color palette of 18 colors

(This is not what the final design will look like, however. Woven designs have MUCH lower resolution than digital images, so a lot of detail will be “lost in translation”.)

Now I needed to find yarns of the right colors to make up the palette. It would essentially be impossible to source these commercially: the colors would have to be nearly precise matches, and the odds of finding them in the limited commercial palette are essentially zero.

To the dye samples!!

I dyed 1500 samples of yarn in various Procion MX color combinations a while back – you can see them here. This gives me an exceptionally large palette to choose from – still doesn’t cover the entire color gamut, but I get quite a bit of selection when I flip through my books.

I was in luck! There were not one, but TWO palettes that looked like they might work:

dye samples compared to original palette

I think I could work with either of these. The bottom set of samples is truer to the colors of the original palette, but (like the original image) that palette doesn’t seem quite Asian enough for me. Too much pink and not enough yellow. The top set of samples has a more yellowish cast, which I think might mimic Asian skin better.

My plan is to dye all six colors, weave up samples, and see which I like best.

I had originally planned to weave using 20/2 silk, which is about the size I wanted, but there’s a hitch: silk dyes differently from cotton. The colors shift, sometimes dramatically. So if I wanted to dye silk, I’d have to do samples to test the dye formulas, adjust as needed, do more samples, etc. Time is short and silk is expensive, so I think I’ve decided to take the quicker path and use a cellulose fiber – probably mercerized cotton.

I’m still debating what size yarn to use, and will likely weave up samples in both 20/2 and 10/2 cotton. There are a lot of technical questions that will need resolving, so I think I’ll start by weaving samples to test yarn sizes. After that I’ll wind and dye my skeins, and start testing colors.

Meanwhile, of course, there’s the electronics to consider. I’m reading through Adafruit’s (lengthy) document about all the different LED products they carry, and am thinking about how to design the cyborg half as well.

And, of course, continuing my powerlifting workouts. In my fantaaaaaaabulous new belt, which arrived last week! I think it’s amazing, and can’t wait to wear it at my meet in San Diego on August 23.

It will go very nicely with my powerlifting competition gear:

Because if you’re going to look like a squashed sausage (“I look good in a wrestling singlet,” said no human ever), you might as well look like a FABULOUS squashed sausage!

July 15, 2024 by Tien Chiu

Weaving and weightlifting

Major progress on the AI weaving project! And also one or two right-angle changes.

I finished weaving the tiger. It came out beautifully:

This design uses four wefts, all brushed mohair/silk yarn (it feels SOOOO yummy!). One weft is orange, one black, and the two remaining wefts are both white (I wanted to get better coverage).

I also wove up a sample with a fifth weft – .75 mm strands of optical fiber! Optical fiber conducts light from end to end with very little light leakage in between – unless, of course, you scratch it up with a knife or sandpaper, in which case the scratched part glows.

(The idea is not original to me. I’d done a piece (of sorts) waaaaaay back when, about ten years ago, in collaboration with Laurie Carlson Steger, who showed me how. She did all the hard parts – I wove an image of a rocket launch and wove in the optical fiber (with her help), but she did all the knife-work to make the piece glow in all the right areas.)

I tested both knife and sandpaper. In the video below, the bright area in the center is scratched with a razor blade, while the bright area on the right is gently sanded with 80 grit sandpaper. That allows the light to leak out wherever I want a glow.

The video shows some of the tricks that you can play with the light. In the video, out of sight, I’m waving a small pocket flashlight over the ends of the strands, which creates a play of light in the glowing spots as the light waxes and wanes over the end of each fiber.

Since each fiber has the potential to be lit up separately from all the others, I’m thinking it would be fun to mount some LEDs at the ends of some optical fibers, to make the design glow or flash or ?? in the specific areas I want. I’m not clear yet on exactly what kind of pattern I want – that depends in part on what can practically be done. So one thing on the list is simply getting a bunch of LEDs and experimenting.

I’m collaborating with a Stanford EE graduate student on this. The plan is for me to do the weaving and Clarissa to do the electronics design and prototyping – it’s going to be really exciting to work with (and hopefully learn from) her.

Now to the right-angle turn in the project. I was noodling on the design of the piece, and what I wanted to say with it. I realized that the tiger, while cool, really didn’t capture the ambiguity I was looking for. Because the ambiguity isn’t between AI and animals, it’s between AI and humans.

Back to DALL-E to get a human face!

After fiddling with (and getting frustrated by) DALL-E for a while, I had an image that seemed like a good start:

I couldn’t seem to convince DALL-E to modify the image to include the entire head – no matter what prompt I used, it wanted to cut off the top of the hair. I did manage to get one image that showed an entire head, though, and the hair was a similar style.

So off to Photoshop and its AI-based Generative Fill. I grafted the top of the other head onto the first image. It didn’t match, of course, so I selected the area and typed into the Generative Fill, “Blend the two photos together seamlessly”. I also flipped the image horizontally.

And here’s what came out:

DALL-E image with top of head added in Photoshop

I’m not going to weave this as a straight-up image. Instead, I’m going to duplicate the left side of the face and put it over the cyborg-y bits, and weave the whole thing as a human face. Then, once the weaving is done, add the circuitry over the top of the face (basically, embroidering it on). I haven’t figured out how I’m going to do the metal bits, as the image is flat. Maybe 3D-print something of the right shape and paint it over? So many possibilities!

The idea behind the piece is the blurred line between AI generated art and human-generated art, and also the blurred line between AI (which often seems human) and humans. I haven’t yet got an official title, but am thinking “Am I AI?” might not be bad.

Still debating materials. I am leaning towards silk weft, both because it will add luster and because 20/2 silk is close to the size I want for the weft. 10/2 cotton (same as the warp) will likely be thicker than I want because I will be working with 4-5 wefts and 2 warps per “pixel” of the design.

Another possibility is a silk/cashmere blend which is either 6000 or 7000 yards per pound, a little thinner than I want, but if I can figure out how to juggle 5-6 shuttles, would also allow me to have more weft colors and hence more nuance in the design. The additional cashmere content might add a little peach-fuzziness to the face, which would be nice.

Still thinking over all the variables!

In other news, I have a new keyboard! I have been having finger issues on my right hand, so my friend Morgan gifted me with a Svalboard Lightly, the successor to the Datahand. It’s one of the most unique keyboards out there, and is completely customizable:

Basically, you put your fingers in the finger wells and type by pressing the center button and/or flicking your fingers up, down, left, right. Thumb has four options: knuckle, nail, finger pad, and up. Those by default are either very commonly used keys (space, enter), keys used in combination with other keys (Control, Alt, etc.) or mode-shift keys (shifting the keyboard from alphabet to numbers/symbols).

You can completely customize the fit by sliding and tilting the finger-wells (and moving the palm rests around), and you can also reprogram the keys to whatever you want. I’ve decided to move away from the traditional QWERTY layout and go with a “modern” configuration called DH Neu, which is designed for ease of typing. I’ve also programmed in some single keys to do common commands, e.g. CTRL-C, CTRL-V, CTRL-X, so I don’t have to do combinations.

I’ve only been using it for a little while, and am still learning to type with it, but it’s WAY more comfortable than any other keyboard I’ve used. It’s not cheap, but you have carpal tunnel or other RSI issues from typing, it’s likely a great tool for you. It has helped people avoid surgery.

I’m in the process of installing two trackballs, one on each side, so I never have to remove my hands from the keyboard to mouse about. Most people use the left trackball to scroll and the right trackball to mouse the cursor around – I’m looking forward to exploring them more. I currently switch between an ergonomic mouse (right hand) and a trackball (left hand) because otherwise I develop RSI problems, so this could be a very efficient and comfortable approach for me.

In powerlifting and health news, I’ve dropped eight pounds since setting a new protein target of 150 grams/day. This is recommended if you want to build muscle (0.8-1 gram of protein for every pound body weight), but it also seems to be making me less hungry, so I’m losing about a pound a month without any effort. This is great! But it’s also put me into a different weight class for competition, which means I will need to re-qualify for Worlds in November. So I’ve signed up for a meet in San Diego in late August – there aren’t any locally that would suit.

And I will be FABULOUS for that meet, of course. I’ve ordered a custom Pioneer belt, with an embroidered design on the back. It’s almost done, and I can’t wait! They sent me a picture of the embroidery a few days ago:

The embroidery will be on the back, and the color will go nicely with my red, yellow, and orange wrestling singlet. Hey, if I’m going to look like a squashed sausage in a skin-tight body suit, at least I’ll look like a FABULOUS sausage. Because it’s not a powerlifting meet, daaahhhhling, it’s a fashion show. LOL!

To close, I’ve been experimenting with ice-dyed tie dyes. You tie the shirt, put it on a grid, put powdered dye and soda ash over it, and then pile ice on top. As the ice melts, the dye seeps into the shirt. The colors can “break” into their component parts, giving really interesting results.

Here’s my favorite so far. It may actually be my favorite of all the tie-dyes I’ve ever made. It looks like batik combined with silk painting, but it’s actually an ice-dyed tie dye. The white lines are created by clamping the folded shirt with curved hemostats (a technique I found on YouTube). I’ll definitely be doing more of these!

Tie-dyed shirt that looks like a batik flower

And, because no blog post would be complete without a cat, here’s the first “take” of the photo. Thanks Fritz!

May 16, 2024 by Tien Chiu

Weaving again (finally!)

I can weave again! Not for long periods yet, but my elbow tendon finally seems to be healing up. I have been weaving in 15-20 minute intervals once or twice a day for the last few days, and all seems to be going well. Thank goodness!

Now that I can weave again, I’ve been creating samples for the AI tiger piece.

I started by dyeing the weft yarns. I’m using a beautiful brushed mohair yarn, with a silk core and kid mohair fluff. Gorgeous stuff, but it only came in white.

I fixed that!

I also dyed a skein of black silk/mohair yarn, but it turned out to be really hard to get a good photo of it. Kinda like Fritz, who turns into a black furry blob when photographed.

After dyeing and rewinding the skeins, I wove the sample blanket for the tiger. Here it is, just off the loom:

What’s a sample blanket? It’s a way of sampling all the colors you plan to use in your piece. First you weave a swatch of each structure you are considering using. Then you take a photo and do a little Photoshop magic to figure out the average color of each swatch, like this:

Next, you go through the swatch collection and remove all the swatches that are essentially duplicates of each other. For example, in the bottom row there are two nearly identical grays and a third one that’s only marginally darker. Removing the duplicates makes it easier to match up colors to weave structures later.

(The white swatch at the very end of the first set is a special, experimental one that I added later. More on that in a bit.)

Once you’ve done that, you tell Photoshop to reduce the image to ONLY those colors found in the swatch set. So with my original image:

gets turned into this image:

The woven piece won’t look exactly like either image, mostly because Photoshop’s assessment of the average color of a swatch doesn’t reflect how the human eye actually sees color. So I don’t worry about minor problems as long as the image overall looks about right.

However, there’s something bothering me in this image, which is that there really isn’t a white. There’s a light-to-medium gray, but I want a brighter white than I can get with the samples I’ve currently woven. I know two different ways to get a lighter color, one that’s truer to white, so I decided to add a “white white” to the swatches. I figured that if it worked, I could explore those options more thoroughly.

And I like the result much better:

Photoshop flattened tiger image with the addition of white.

(If you’re wondering where the cyborg half of the tiger went, by the way, I’ll be adding circuit boards and “circuitry” to the right side after weaving. I wanted to weave the entire white-tiger half to give me the most options for adding circuitry later.)

I still have to weave more samples and noodle on exactly how I’m going to achieve that white, but I’m definitely making progress!

In other news, I’m getting ready to harvest about 60 bulbs of garlic. I decided to rotate out tomatoes from my larger garden plot this year, so I planted three varieties of garlic instead, and am only growing eight or nine tomato plants in the small plot. That includes one F1 cross (hybrid) that I think will produce some verrrry interesting results next year, and also includes a few more tomatoes that I’d like to breed. So tomatoes are minimal this year, but will be back in full force next year. And I expect to be giving away a lot of garlic soon!

Once the garlic’s harvested, I’ll plant sweet potatoes. We have a super long growing season (it frosts once or twice in January and that’s it) so I think I can get sweet potatoes this year, even with a late start.

And our roses are going gangbusters. Here’s just one bouquet of the fragrant beauties.

Finally, no blog post would be complete without a cat. Or, in this case, two cats.

Fritz and Tigress are just about eleven years old now (time flies!), and they’ve definitely slowed down, but they are still very much the rulers of the household. I spend all my free time brushing Tigress when she demands it (about eighteen times a day), giving Fritz belly rubs when he wants them (about twenty-six times a day), and generally catering to feline demands. Such is the life of a human!

March 26, 2024 by Tien Chiu

Exploring some new ideas

I had a really interesting conversation a week or two ago with Hideo Mabuchi. Hideo is at Stanford, where he is a professor of applied physics, ceramicist, faculty head of the Stanford Arts Institute, and (most relevantly from my perspective) a weaver. He recently got a TC-2 loom for Stanford, and since Stanford, California College of the Arts, and I are the the only people I know of in the Bay Area with a TC-2, I thought we ought to meet.

Hideo is generally interested in bridging the divide between STEM and art/craft, so we talked a bit about his plans for their TC-2. He had a whole bunch of fascinating ideas about innovative uses of a TC-2 combined with technology – which I found frankly much more personally compelling than trying to figure out how to weave interesting imagery. I am more of a researcher than an artist (not surprising given that both my parents were scientists and I majored in math at Caltech), so my brain is now full of all sorts of crazy ideas, most of which will require help from someone on the STEM side to implement.

The first step, though, is doing the AI tiger project. Hideo helped connect me with a first-year EE PhD student who’s interested in helping out with the electronics portion of the tiger. She’s away on spring break, but once she’s back we’ll meet up and talk about possible ways to implement the “cyborg” half of the tiger. It will be fun to collaborate with her!

I also have, stewing in the back of my mind, an urge to weave a book. A flag book, to be precise. Here’s an example, photographed by Jimmie Quick:

The basic idea is an accordion spine with tabs pasted onto the spine in one direction, then (below) in the other. Two sets of flags is the minimum, but often there are a lot more flags.

I think it should be possible to weave these in double weave, perhaps using a shrinking weft for the spine, to make a woven origami book. I’m not 100% certain it will work but it’s certainly worth a try. If I can do the pages in double weave I could put inserts into each “page,” which would open up all sorts of possibilities. I could do pockets and put imagery, coins, ?? into each pocket!

I have noticed that I am generally more interested in figuring out techniques and structures than in creating something with a message (i.e., art). I had aspirations towards being an Artist at some point, but frankly, I think the kind of digging I’m currently doing is much more interesting. I look forward to exploring more STEM crossovers and intriguing three-dimensional structures!

In powerlifting news, I started with a new coach two and a half weeks ago. Her name is Leah, and she’s a head coach with Barbell Medicine. They have a verrrry interesting podcast where once a month or so they dive in-depth into a weightlifting-related topic or issue. They read through all the medical research available, then offer their conclusions based on actual study results, with citations. In short, they are a font of science-based, evidence-based powerlifting expertise. That is extremely rare in a sport where most training is pretty idiosyncratic.

Anyway, Leah’s been coaching powerlifters for nine years and is a great powerlifter herself, competing both nationally and at Worlds. Just as importantly, she has worked with quite a few older lifters. Most powerlifting trainers are young(ish) men coaching young men. Nothing wrong with that, but a 53-year-old woman is going to have different training concerns than a 23-year-old man. In particular, I have a higher likelihood of injury and a slower recovery time than young’uns, so it’s good to have a coach who can help me find the right training cadence and intensity.

I’m VERY happy with how training is going so far. It’s very different from anything I’ve done earlier – the main difference being that every session is calibrated based on how much I can lift that day – as opposed to yesterday, or at the last meet, or my all-time record. So if I’m having a bad day and can’t lift as much as usual, then the weights I’m lifting get scaled back. If I’m having a good day, heavier weights are in order. This means I’m always lifting at an appropriately challenging weight, and vastly reduces the chances of injury. TOTALLY psyched for this!

I also realized that I’ve come a looong way in the 5.5 years since I began powerlifting. I’ve more than tripled my strength (85-95 lb squat as a beginner, 310 lb today), lost about 4 inches off my waist, and gotten massively more muscular. I’ve even gone up a bra size – not because my boobs have changed size but because my back and shoulder muscles are now enormous!

While I haven’t lost an ounce, I’m pretty sure I’ve exchanged at least 20 pounds of fat for muscle. I’ll never be a sports car, but I’m rather enjoying being a bulldozer.

My next powerlifting meet is Saturday, May 4. Five and a half weeks out. I’m looking forward to this.

Meanwhile, from last week’s workout, three reps of deadlift at 245 lbs. Easy as pie.

February 19, 2024 by Tien Chiu

Fire in the rain

Well, the velvet attempt was a bit of a disaster. I’m still determined to weave velvet, but I’ve realized that debugging the process and solving all the problems is likely to take months. More on that in a future blog post. (I haven’t given up, just contemplating my next move.)

Since I would really like to have a finished project – any project! – someday, I’ve decided to warp up the left side of Grace. I was initially thinking about a black and white warp, which is what most people put onto a jacquard loom because of its versatility. However, black and white is booooooring. (I also almost never use white in my designs.) So I decided to create another “fire” warp instead.

Here’s the one I cut off Maryam when I sold her:

warp dyed in many shades of orange, yellow, and red

It’s a mix of different fibers and yarn sizes. Here’s how that warp looked, before dyeing:

warp prior to dyeing, showing the different types of yarn in the warp

I think that warp was a mix of mercerized cotton, silk, and unmercerized cotton. The fibers took the dye differently, producing the gorgeous variegation shown above.

This new warp is a different mix. It’s one strand of 10/2 tencel, one strand of mercerized 10/2 cotton, one strand of a 4/28 nm cabled silk, and one strand of 20/2 silk. They should absorb dyes differently, and also reflect light differently – some are very lustrous, others are not. It should be interesting!

Here’s the warp halfway through winding:

partially wound warp on a large warping mill

I wound the entire 29″ warp in a single bout. I know you’re not supposed to do this, but with painted warps, winding multiple bouts usually results in abrupt color shifts in the middle of the warp. There are ways to make that less noticeable, but for me, it’s easier just to wind a super-wide warp, dye it, and then separate it out into smaller sections when warping.

(I can get away with winding a very wide warp because I’ve got a large warping mill. I think this approach would not work nearly as well on a warping board.)

Now to the rain part. It’s been pouring about half the time, which is fantastic because this is California and we (always) desperately need rain. But my dye studio is on the patio, and, well, it’s raining for the next four days. (Hurray! Boo! Hurray!)

Fortunately, we have a table that sits under an underhang, and stays dry. So I sneaked out during a lull in the rain and did a little dyeing:

It looks like a solid color in the photo, but once it’s rinsed and dried, it should be all shades of fire. (Crossing fingers that this actually happens – dyeing always entails some randomness.)

I used 1.75 gallons (really!) of golden yellow to start, pouring it over the warp and squishing to soak the entire warp in yellow. (I really didn’t want white spots.) Then I drizzled 2 quarts of orange and red-orange over it, flipped it over, and squirted more orange/red on the yellow spots. My goal was a mostly orange warp with some areas of red-orange and a few areas of gold.

Tomorrow morning, I’ll start rinsing out the warp. I think it will work great for my AI-generated tiger, so I’m going to spend the next week or two working a bit more on the design. In particular, I want to think about how I want to render the circuitry – whether to weave it in metallic yarns, embroider it, couch real circuit boards down, or (this sounds exciting) use the retro-reflective yarns I bought ages ago to create brilliantly reflective areas.

(Retro-reflective yarn, in case you’re wondering, is a plastic yarn that’s been coated with the same glass beads that are used for high-visibility/reflective clothing. It looks flat gray until you shine a light directly at it, at which point it turns brilliantly white with reflected light.)

In case you’d forgotten, here’s the tiger I plan to weave:

The left part of the tiger will use the Fire warp, and the right side will use mostly the black warp, I think.

Can’t wait to see how this all turns out!