Hector retweeted
ComfyUI experiment #1
Currently testing this Text2Animation from HY-Motion. It's an interesting approach and still requires some cleanup. Took 30 seconds to generate a 12 seconds animation. Then got it retargeted for UEFN characters.
I have a big list of ComfyUI tests I want to do. Maybe 360 environments So far i like to run this stuff locally on my GPU.
Hector retweeted
✨We are excited to open-source Tencent HY-Motion 1.0, a billion-parameter text-to-motion model built on the Diffusion Transformer (DiT) architecture and flow matching. Tencent HY-Motion 1.0 empowers developers and individual creators alike by transforming natural language into high-fidelity, fluid, and diverse 3D character animations, delivering exceptional instruction-following capabilities across a broad range of categories. The generated 3D animation assets can be seamlessly integrated into typical 3D animation pipelines.🎮🎥
Highlights:
🔹Billion-Scale DiT: Successfully scaled flow-matching DiT to 1B+ parameters, setting a new ceiling for instruction-following capability and generated motion quality.
🔹Full-Stage Training Strategy: The industry’s first motion generation model featuring a complete Pre-training → SFT → RL loop to optimize physical plausibility and semantic accuracy.
🔹Comprehensive Category Coverage: Features 200+ motion categories across 6 major classes—the most comprehensive in the industry, curated via a meticulous data pipeline.
🌐Project Page: hunyuan.tencent.com/motion
🔗Github: github.com/Tencent-Hunyuan/H…
🤗Hugging Face: huggingface.co/tencent/HY-Mo…
📄Technical report: arxiv.org/pdf/2512.23464
Evident what @UnrealEngine Metahuman Creator's greatest pitfalls, bad eye lash/eye shadows, jaw problems in east asian females, hair groom issues, skin textures
what if you combined Unreal Engine, MetaHumans, and the latest Generative AI models?
Watch the full video: youtu.be/uhJR4S8qj88?si=pNVR…
Hector retweeted
what if you combined Unreal Engine, MetaHumans, and the latest Generative AI models?
Watch the full video: youtu.be/uhJR4S8qj88?si=pNVR…
Hector retweeted
How @Netflix got mad because I wouldn’t hand over my prompts for FREE - A thread ⤵️
A Netflix AI division asked me to do a free test for a potential job. I usually refuse free tests because free doesn't pay bills, but being Netflix I said why not, it may be a good opportunity.
Hector retweeted
Introducing SAM 3D, the newest addition to the SAM collection, bringing common sense 3D understanding of everyday images. SAM 3D includes two models:
🛋️ SAM 3D Objects for object and scene reconstruction
🧑🤝🧑 SAM 3D Body for human pose and shape estimation
Both models achieve state-of-the-art performance transforming static 2D images into vivid, accurate reconstructions.
🔗 Learn more: go.meta.me/305985
Exciting new: Meta releases segment (SAM3D-BODY + MHR) Seems quite compatible with #UnrealEngine 5.7 parametric #Metahuman ! Can imagine a workflow leveraging this tech; Image/Prompt -> GLB -> Unreal Engine Pose/animation asset!
github.com/facebookresearch/…
github.com/facebookresearch/…
Interesting strategy. So if strategy markets dumps their btc OTC, then buys up etc at firesale price, they could maybe 1.5x their btc stack
DAT Mechanics Part 2
Now let’s consider the case where DATs have varying levels of leverage. In this case, a DAT levered or not could grief the other DATs of the same underlying by outright dumping the full underlying stash at the right time creating margin calls/cascades for everyone else and the scoop up the cheap underlying once everyone is forced to puked. While the initial DAT is out of the underlying, they are fully backing their stock price (net of slippage) with cash while everyone else gets washed out. This play is possible when the DATs have no leverage but it is harder to pull off.
Now consider the case where a DAT holds the locked version of an underlying which it acquired either from PIPE direct contributions or purchased from large holders from PIPE cash contributions. If the discount on locked underlying is high, this is good for DAT equity shareholders, otherwise, it is bad for shareholders. Here we have to be precise about what mNAV means. mNAV only makes sense when we consider the underlying NAV in the denominator with a discount for lockups. But what is the market clearing discount? This is tricky but we can look at OTC markets for some data points. We could also do things in reverse and look at different DAT market caps after some initial trading and infer the market discount on the underlying by comparing them. Now if we consider arb dynamics against true mNAV, we see the same analysis for liquid underlyings as we see for locked underlyings with one exception which is that the arb is harder to close because the locked underlying market is just less liquid.
Next, there are also considerations for certain underlyings where the foundation has veto rights against the DAT for the selling of the underlying. Obviously in this case, mNAV can go lower for that DAT before arbs are taken. If there are competitor DATs, this bodes well for them and badly for the DAT that gets constantly vetoed.
Lastly, there are correlation effects between different underlyings for obvious reasons.
So far we have been focusing on underlying compression. What about underlying expansion? Well one thing that causes underlying expansion is the creation of new DATs for the underlying. Generally new DATs are created when mNAVs of competitors are high. For many underlyings, this is no longer the case so we can expect the rate of new DAT creations to decrease with time.
What about a DAT like MSTR getting included in an index. Well this is obviously good for the underlying because price insensitive buying of the equity will increase mNAV giving more cushion for more ATMs and hence more price insensitive buying of the underlying.
What about DATs that eventually create an operating business using their balance sheet of the underlying like a lending business. In the long term, this is good for the industry because any true bull run requires credit expansion. After the 3AC/Genesis/BlockFi credit crisis, we never truly had the same level of circulating credit which is why this last bull run has felt so cynical and muted. So the return of lending desks from within DATs is a good sign for what’s to come in the long run. There’s a corollary here that it actually makes sense to have a few stablecoin DATs. Sounds bad but is actually very good. It serves as a counterweight for credit expansion against people who borrow just to short underlyings. With stablecoin DATs turned lending desk, you can borrow to long underlyings.
In the short run, without stablecoin DATs to lend to longs, most borrows from DATs turned lending desks will be for shorting. If the underlying has other uses besides just shorting then the downward effect will be reduced.
Last thing I will say is that DAT manager skill is important. Each underlying will have winner-take-most dynamics for its DATs.
Hector retweeted
DAT Mechanics Part 2
Now let’s consider the case where DATs have varying levels of leverage. In this case, a DAT levered or not could grief the other DATs of the same underlying by outright dumping the full underlying stash at the right time creating margin calls/cascades for everyone else and the scoop up the cheap underlying once everyone is forced to puked. While the initial DAT is out of the underlying, they are fully backing their stock price (net of slippage) with cash while everyone else gets washed out. This play is possible when the DATs have no leverage but it is harder to pull off.
Now consider the case where a DAT holds the locked version of an underlying which it acquired either from PIPE direct contributions or purchased from large holders from PIPE cash contributions. If the discount on locked underlying is high, this is good for DAT equity shareholders, otherwise, it is bad for shareholders. Here we have to be precise about what mNAV means. mNAV only makes sense when we consider the underlying NAV in the denominator with a discount for lockups. But what is the market clearing discount? This is tricky but we can look at OTC markets for some data points. We could also do things in reverse and look at different DAT market caps after some initial trading and infer the market discount on the underlying by comparing them. Now if we consider arb dynamics against true mNAV, we see the same analysis for liquid underlyings as we see for locked underlyings with one exception which is that the arb is harder to close because the locked underlying market is just less liquid.
Next, there are also considerations for certain underlyings where the foundation has veto rights against the DAT for the selling of the underlying. Obviously in this case, mNAV can go lower for that DAT before arbs are taken. If there are competitor DATs, this bodes well for them and badly for the DAT that gets constantly vetoed.
Lastly, there are correlation effects between different underlyings for obvious reasons.
So far we have been focusing on underlying compression. What about underlying expansion? Well one thing that causes underlying expansion is the creation of new DATs for the underlying. Generally new DATs are created when mNAVs of competitors are high. For many underlyings, this is no longer the case so we can expect the rate of new DAT creations to decrease with time.
What about a DAT like MSTR getting included in an index. Well this is obviously good for the underlying because price insensitive buying of the equity will increase mNAV giving more cushion for more ATMs and hence more price insensitive buying of the underlying.
What about DATs that eventually create an operating business using their balance sheet of the underlying like a lending business. In the long term, this is good for the industry because any true bull run requires credit expansion. After the 3AC/Genesis/BlockFi credit crisis, we never truly had the same level of circulating credit which is why this last bull run has felt so cynical and muted. So the return of lending desks from within DATs is a good sign for what’s to come in the long run. There’s a corollary here that it actually makes sense to have a few stablecoin DATs. Sounds bad but is actually very good. It serves as a counterweight for credit expansion against people who borrow just to short underlyings. With stablecoin DATs turned lending desk, you can borrow to long underlyings.
In the short run, without stablecoin DATs to lend to longs, most borrows from DATs turned lending desks will be for shorting. If the underlying has other uses besides just shorting then the downward effect will be reduced.
Last thing I will say is that DAT manager skill is important. Each underlying will have winner-take-most dynamics for its DATs.
DAT Mechanics Part 1
Some folks saying this the next LUNA (it’s not). Others are saying there’s nothing to unwind (there is). Think carefully about the mechanics.
A lot of these DATs have no leverage so there won’t be any forced selling from getting called on debt. Still there is an incentive to sell the underlying and buy the stock back to capture the arb and push mNAV up. But how much of an incentive is there and how far up should mNAV be. This largely depends on the underlying asset. In the case of BTC, there is one very large player (MSTR) and many much smaller players. So you could say that MSTR alone as a DAT drives BTC price. Since mNAV here is healthy and since the leverage is controlled well, you probably won’t see much selling of the underlying here. As a side note, it makes sense that the debt is at a safe level since Saylor got such a great deal against the buyers of the early converts since he was the only game in town selling crypto upside to fixed income desks.
Now imagine an underlying where the distribution of DAT holdings isn’t as extremely skewed toward one mega holder. If there is a small handful of major DATs for a single underlying, then this becomes a Prisoner Dilemma coordination game where the good outcome is possible to some extent. At .9 or .95 mNAV, it is conceivable that none of the players go for the arb. Even without outright coordination, they could all choose to (3,3). Now suppose mNAV drops to .5 as an extreme example. Now the arb profits are much bigger and could outweigh the negative consequences of underlying market cap compression, reflexive selling effects, and reputational considerations. So we can see that whether there is an underlying unwind is dependent on a whole host of factors including how many large players are running DATs on that underlying. In other words, the gain in arb value is individual while the compression of the underlying market cap and hence all DAT market caps for the same underlying is socialized. All else equal, more big DATs is worse than fewer big DATs for this game.
As a tangent, I’m roughly assuming above that most DATs for the same underlying have similar mNAVs. If this wasn’t the case, we could see 1) different levels of incentives to do the arb and 2) one DAT buying another DATs equity. An exercise left to the reader.
Now let’s introduce more complexity into this toy model. Each manager of a DAT for the same underlying has a different propensity to take the arb. It depends how crypto native vs tradfi native the manager is. Crypto natives don’t want the music to slow. For tradfi natives, well, whoever sells first sells best. Also there are reputational considerations. Whoever is more prominent has more personal considerations outside of managing the DAT. These considerations could push up or down the threshold for where arbs will be taken.
These people miss the point. They aren't building on eth, but their L2, whether it be @Optimism or @arbitrum.
Hector retweeted
DotA is a gentleman’s game. LoL is for children.
Looks like a noctua nh d15s CPU cooler 🤣🤣
The Persian wind tower (بادگیر) or how a 700-year-old air conditioner could cool an environment up to 12°C with no electricity
[📹 Never Enough Architecture]
teapawt.org/i/status/1780932805926…