EMO: Alibaba’s Diffusion Model-Based Talking Portrait Generator

Alibaba’s EMO (or Emote Portrait Alive) framework is a recent entry in a series of attempts to generate a talking head using existing audio (spoken word or vocal audio) and a reference portrait image as inputs. At its core it uses a diffusion model that is trained on 250 hours of video footage and over 150 million images. But unlike previous attempts, it adds what the researchers call a speed controller and a face region controller. These serve to stabilize the generated frames, along with an additional module to stop the diffusion model from outputting frames that feature a result too distinct from the reference image used as input.

In the related paper by [Linrui Tian] and colleagues a number of comparisons are shown between EMO and other frameworks, claiming significant improvements over these. A number of examples of talking and singing heads generated using this framework are provided by the researchers, which gives some idea of what are probably the ‘best case’ outputs. With some examples, like [Leslie Cheung Kwok Wing] singing ‘Unconditional‘ big glitches are obvious and there’s a definite mismatch between the vocal track and facial motions. Despite this, it’s quite impressive, especially with fairly realistic movement of the head including blinking of the eyes.

Meanwhile some seem extremely impressed, such as in a recent video by [Matthew Berman] on EMO where he states that Alibaba releasing this framework to the public might be ‘too dangerous’. The level-headed folks over at PetaPixel however also note the obvious visual imperfections that are a dead give-away for this kind of generative technology. Much like other diffusion model-based generators, it would seem that EMO is still very much stuck in the uncanny valley, with no clear path to becoming a real human yet.

Continue reading “EMO: Alibaba’s Diffusion Model-Based Talking Portrait Generator”

A business card-sized love detector in a 3D-printed package.

2024 Business Card Challenge: Who Do You Love?

When you hand your new acquaintance one of your cards, there’s a chance you might feel an instant connection. But what if you could know almost instantly whether they felt the same way? With the Dr. Love card, you can erase all doubt.

As you may have guessed, the card uses Galvanic Skin Response. That’s the fancy term for the fact that your skin’s electrical properties change when you sweat, making it easier for electricity to pass through it. There are two sensors, one on each short end of the card where you would both naturally touch it upon exchange. Except this time, if you want to test the waters, you’ll have to wait 10-15 seconds while Dr. Love assesses your chemistry.

The doctor in this case is an RP2040-LCD-0.96, which is what it sounds like — a Raspberry Pi Pico with a small LCD attached. For the sensors, [Un Kyu Lee] simply used 8mm-wide strips of nickel. If you want to build your own, be sure to check out the build guide and watch the video after the break for a demonstration of Dr. Love in action.

Continue reading “2024 Business Card Challenge: Who Do You Love?”

A Peek Inside Apple Durability Testing Labs

Apple is well-known for its secrecy, which is understandable given the high stakes in the high-end mobile phone industry. It’s interesting to get a glimpse inside its durability labs and see the equipment and processes it uses to support its IP68 ingress claims, determine drop ability, and perform accelerated wear and tear testing.

Check out these cool custom-built machines on display! They verify designs against a sliding scale of water ingress tests. At the bottom end is IPx4 for a light shower, but basically no pressure. Next up is IPx5, which covers low-pressure ambient-temperature spray jets from all angles – we really liked this machine! Finally, the top-end IPx7 and IPx8 are tested with a literal fire hose blast and a dip in a static pressure tank, simulating a significant depth of water. An Epson robot arm with a custom gripper is programmed to perform a spinning drop onto a hard surface in a repeatable manner. The drop surface is swapped out for each run – anything from a wooden sheet to a slab of asphalt can be tried. High-speed cameras record the motion in enough detail to resolve the vibrations of the titanium shell upon impact!

Accelerated wear and tear testing is carried out using a shake table, which can be adjusted to match the specific frequencies of a car engine or a subway train. Additionally, there’s an interview with the head of Apple’s hardware division discussing the tradeoffs between repairability and durability. He makes some good points that suggest if modern phones are more reliable and have fewer failures, then durability can be prioritized in the design, as long as the battery can still be replaced.

The repairability debate has been raging strong for many years now. Here’s our guide to the responsible use of new technology.

Continue reading “A Peek Inside Apple Durability Testing Labs”

8-Tracks Are Back? They Are In My House

What was the worst thing about the 70s? Some might say the oil crisis, inflation, or even disco. Others might tell you it was 8-track tapes, no matter what was on them. I’ve heard that the side of the road was littered with dead 8-tracks. But for a while, they were the only practical way to have music in the car that didn’t come from the AM/FM radio.

If you know me at all, you know that I can’t live without music. I’m always trying to expand my collection by any means necessary, and that includes any format I can play at home. Until recently, that list included vinyl, cassettes, mini-discs, and CDs. I had an 8-track player about 20 years ago — a portable Toyo that stopped working or something. Since then, I’ve wanted another one so I can collect tapes again. Only this time around, I’m trying to do it right by cleaning and restoring them instead of just shoving them in the player willy-nilly.

Update: I Found a Player

A small 8-track player and equally small speakers, plus a stack of VHS tapes.
I have since cleaned it.

A couple of weeks ago, I was at an estate sale and I found a little stereo component player and speakers. There was no receiver in sight. I tested the player with the speakers and bought them for $15 total because it was 75% off day and they were overpriced originally. While I was still at the sale, I hooked it up to the little speakers and made sure it played and changed programs.

Well, I got it home and it no longer made sound or changed programs. I thought about the play head inside and how dirty it must be, based on the smoker residue on the front plate of the player. Sure enough, I blackened a few Q-tips and it started playing sweet tunes again. This is when I figured out it wouldn’t change programs anymore.

I found I couldn’t get very far into the player, but I was able to squirt some contact cleaner into the program selector switch. After many more desperate button presses, it finally started changing programs again. Hooray!

I feel I got lucky. If you want to read about an 8-track player teardown, check out Jenny List’s awesome article. Continue reading “8-Tracks Are Back? They Are In My House”

Comparing Those Ten Cent Microcontrollers

If you follow the world of microcontrollers, then you’ll probably be familiar with the most recent crop of ten cent parts. They bring power and features previously the preserve of much more expensive chips into the super-budget arena, and they’re appearing in plenty of projects on these pages.

If you’re not familiar with them it can seem daunting to decide which one to use, so to help you [Zach of All Trades] is comparing two of the more common ones. The CH32V003 with a RISC-V core and the PY32F002 with an ARM Cortex M0+ core are both pretty similar on paper, but which should you use?

The video below gives a run-down of each part along with some demonstrations before making its conclusions. The ARM-based part isn’t as quick as the RISC-V one but has a slight edge on peripherals, while the support is where a potential winner emerges in the shape of the CH32. That should be the last word, but for that the PY32 has the distance advantage over its rival of ready availability.

So this look at two families of cheap microcontrollers reveals the pros and cons of each, but in reality it provides an introduction to two sets of powerful chips for pennies.

As we’ve observed before, there are more chips to be found in this market.

Continue reading “Comparing Those Ten Cent Microcontrollers”

A LEGO Orrery

We aren’t sure how accurate you can get with LEGO, but a building block orrery looks cool, if nothing else. [Marian42] saw one done a few years ago and decided to build a version with a different mechanism. At first, the plan was to use some 3D printed fixtures, but the final product is made entirely from LEGO bricks. Very impressive. The video below shows that it has been complete for awhile, but the write-up that goes into great detail has only just arrived and it was worth the wait.

This is one of those things that seems simple if you don’t think too hard about it. However, when you sit down to actually do it, there are a number of challenges. For one thing, the Earth tilts at 23.5 degrees, and as the planet rotates, the tilt stays in the same direction, making it tricky to model mechanically.

The moon also has a 5.15 degree inclination, but since that’s hard to notice at this scale, the LEGO orrery exaggerates it. So, the Moon’s track has its own set of design problems. The whole thing has to rotate on a concentric shaft, which is also tricky to get right with kids’ building blocks.

Compared to the last orrery we saw, this one is huge. We’ve always been partial to ones that you have to look up to.

Continue reading “A LEGO Orrery”

Can A Toy Printer Be Made Great?

Now that the bottom end of the 3D printer market has been largely cleared of those garbage “Prusa i3 clone” models which used to infest it a few years ago, a new breed of ultra-cheap printer has taken their place. EasyThreed make a range of very small printers pitched as toys, and while they’re no great shakes by the standards of most Hackaday readers, they do at least work out of the box. For their roughly $75 price tag they deliver what you’d expect, but can such a basic machine be improved with a few upgrades? [Made with Layers] has taken a look.

These printers have an all-plastic snap-together construction with a 10 cm by 10 cm bed and a set of small geared stepper motors driving their axes. He concentrates on stiffening the structure, upgrading those motors, and because he’s sponsored by a 3D printer electronics company, upgrading their controller.

The motors were replaced first with some NEMA 11 steppers, and then by some over-sized ones which maybe push the idea a little far. By moving the motors to a bracket he was able to free up their mountings to secure a 3D printed insert to stiffen the arms. Perhaps he’s pushing it a little for the video with the electronics upgrade, but we think there’s a happy medium with the smaller of the two motor upgrades and the stiffening.

So if you have an EasyThreed in your life it’s possible to upgrade it into something a little better, but it’s worth asking whether that $75 might be better spent in saving for a better machine in the first place. We’ve been curious about these tiny printers for a while though, and it’s interesting to have some more of our questions answered.

Continue reading “Can A Toy Printer Be Made Great?”