Загрузка видео...

Не удалось загрузить видео

На главную

Many generative models work in an autoencoder's latent space; They train faster and generate better when this space is well organized. One way to organize it is distilling a pretrained model into it. But how can it work when the two spaces have different structures?

kyutai_labs's profile picture

kyutai

27,836 subscribers

11,748 просмотров • 10 дней назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

A pergola can do more than define an outdoor space—it can influence the microclimate around it. In summer, the leafy canopy provides shade and reduces heat gain, creating a cooler, more comfortable outdoor space. In winter, when the vegetation is less dense, more sunlight can reach the space, allowing it to warm naturally. It’s a simple example of how landscape, architecture, and seasonal design can work together to create comfort.
0:10

A pergola can do more than define an outdoor space—it can influence the microclimate around it. In summer, the leafy canopy provides shade and reduces heat gain, creating a cooler, more comfortable outdoor space. In winter, when the vegetation is less dense, more sunlight can reach the space, allowing it to warm naturally. It’s a simple example of how landscape, architecture, and seasonal design can work together to create comfort.

Architecture Presentation

19,793 просмотров • 16 дней назад

who is behind space bunny? it’s been a week since this stealth model went live and been testing it daily. compared to day one, it’s now: > way faster > giving way better outputs > still free for now opencode extended it for another week and looks like latest checkpoint got huge update you can make animations like opus 5.5 with space bunny now and made this one with it, not bad for a free model i think official model will drop next week, but still can’t figure out who is behind it lmaoo
0:15

who is behind space bunny? it’s been a week since this stealth model went live and been testing it daily. compared to day one, it’s now: > way faster > giving way better outputs > still free for now opencode extended it for another week and looks like latest checkpoint got huge update you can make animations like opus 5.5 with space bunny now and made this one with it, not bad for a free model i think official model will drop next week, but still can’t figure out who is behind it lmaoo

J A Z I I

18,343 просмотров • 12 дней назад

Kedarnath is a tremendous space. The utterance of the sound “Shiva” attains a completely new dimension and significance in Kedar. It is a space which has been specially prepared for this particular sound. When we utter the word “Shiva,” it is the freedom of the uncreated, the liberation of one who is not created. It is almost like on this planet, the sound “Shiva” emanates from this place. For thousands of years, people have experienced that space as a reverberation of that sound. This is also a place that has witnessed thousands of Yogis and mystics of every kind. When I say every kind, you cannot imagine those kinds. These are people who made no attempt to teach anything to anyone. Their way of making an offering to the world was by leaving their energies, their path, their work – everything – in a certain way in these spaces.
0:13

Kedarnath is a tremendous space. The utterance of the sound “Shiva” attains a completely new dimension and significance in Kedar. It is a space which has been specially prepared for this particular sound. When we utter the word “Shiva,” it is the freedom of the uncreated, the liberation of one who is not created. It is almost like on this planet, the sound “Shiva” emanates from this place. For thousands of years, people have experienced that space as a reverberation of that sound. This is also a place that has witnessed thousands of Yogis and mystics of every kind. When I say every kind, you cannot imagine those kinds. These are people who made no attempt to teach anything to anyone. Their way of making an offering to the world was by leaving their energies, their path, their work – everything – in a certain way in these spaces.

Sadhguru

57,872 просмотров • 5 месяцев назад

Wolves trying Joan Beringer in a two big lineup in the summer. Watching him work through it. Rim runs and dives and then gets back out to space. Sees his man show help and cuts (even though a roll is happening). They work it out, he goes to work in the post.
0:16

Wolves trying Joan Beringer in a two big lineup in the summer. Watching him work through it. Rim runs and dives and then gets back out to space. Sees his man show help and cuts (even though a roll is happening). They work it out, he goes to work in the post.

Steve Jones

44,925 просмотров • 3 месяцев назад

SECRETARY RUBIO: The only way Cuba can have a better future is if it has a different economic model.
0:12

SECRETARY RUBIO: The only way Cuba can have a better future is if it has a different economic model.

Department of State

101,028 просмотров • 7 месяцев назад

Introducing Turbodiff This is the fastest way to review a PR Using SOTA speed reading technology you can review your agent’s work faster than they can write it Try it today at
0:15

Introducing Turbodiff This is the fastest way to review a PR Using SOTA speed reading technology you can review your agent’s work faster than they can write it Try it today at

jacob paris ▲

55,513 просмотров • 3 месяцев назад

Tsundere Malfoid slop will continue until morale improves. One thing I'm using is a 3D latent upscaling with minimax H3... and one of the great things is that it's faster to get high-resolution videos. I can generate quickly at low resolution and then upscale it to finish. The main problem is that it often garbles the text, so if I try anything with Japanese audio and English subtitles, it doesn't work. So in those cases, I have to generate at a relatively high resolution to begin with, which takes longer.
0:15

Tsundere Malfoid slop will continue until morale improves. One thing I'm using is a 3D latent upscaling with minimax H3... and one of the great things is that it's faster to get high-resolution videos. I can generate quickly at low resolution and then upscale it to finish. The main problem is that it often garbles the text, so if I try anything with Japanese audio and English subtitles, it doesn't work. So in those cases, I have to generate at a relatively high resolution to begin with, which takes longer.

Emerald Apple

37,310 просмотров • 5 дней назад

These have been great when they work but they do dumb things more often than not. This one was not docking correctly so I reset it to the charging station and told it to go recharge. This is what it proceeded to do. Tech is great but it’s not ready yet.
0:13

These have been great when they work but they do dumb things more often than not. This one was not docking correctly so I reset it to the charging station and told it to go recharge. This is what it proceeded to do. Tech is great but it’s not ready yet.

John Kaminski, Ph.D.

20,795 просмотров • 3 лет назад

Been wanting to film one of these because it happens so consistently and highlights a subtle but significant part of the V14 brain. Not only does V14 recognize collapsing lanes way sooner than V13 when it is in them and needs to get out, it also recognizes them when it is not in them, and is aware that vehicles inside of them need to get over. Here’s one of the many times I’ve witnessed V14 keeping space open for cars it knows will have to switch lanes.
0:12

Been wanting to film one of these because it happens so consistently and highlights a subtle but significant part of the V14 brain. Not only does V14 recognize collapsing lanes way sooner than V13 when it is in them and needs to get out, it also recognizes them when it is not in them, and is aware that vehicles inside of them need to get over. Here’s one of the many times I’ve witnessed V14 keeping space open for cars it knows will have to switch lanes.

Mike P

28,008 просмотров • 10 месяцев назад

Perplexity Computer is massively multi-model. Computer orchestrates models to run agents in parallel, leveraging Opus to match each task to the model best suited for it. In total, Computer can route work across 19 different models.
0:12

Perplexity Computer is massively multi-model. Computer orchestrates models to run agents in parallel, leveraging Opus to match each task to the model best suited for it. In total, Computer can route work across 19 different models.

Perplexity

534,569 просмотров • 7 месяцев назад

WOW: 🇯🇵 Japan is now one of the first countries ever to generate electricity in space and send it back to earth!
0:14

WOW: 🇯🇵 Japan is now one of the first countries ever to generate electricity in space and send it back to earth!

Coinvo

115,492 просмотров • 4 месяцев назад

This is actually true. Women are happiest when they don't need to think. They just arrive at an airport and a jet is waiting. How does a private jet get organized? Who knows. New country, arrive at a 5-star, have a great time. Have no idea how they got there, how much it cost, how it was organized, how they're getting home. Zero. They literally want to turn their brains off and smile when they're with the man they love. Logistics kill smiles.
0:13

This is actually true. Women are happiest when they don't need to think. They just arrive at an airport and a jet is waiting. How does a private jet get organized? Who knows. New country, arrive at a 5-star, have a great time. Have no idea how they got there, how much it cost, how it was organized, how they're getting home. Zero. They literally want to turn their brains off and smile when they're with the man they love. Logistics kill smiles.

Andrew Tate

7,501,404 просмотров • 3 лет назад

This animation shows how a uniform grid of points is deformed by a flow matching model. The visualization highlights how the learned flow warps the underlying space—stretching, compressing, and bending it—to transform one probability distribution into another.
0:13

This animation shows how a uniform grid of points is deformed by a flow matching model. The visualization highlights how the learned flow warps the underlying space—stretching, compressing, and bending it—to transform one probability distribution into another.

Alec Helbling

43,008 просмотров • 1 год назад

Okay this new model is insane. Generate an image, then give it to the model to turn it into a 3D model with matching texture in seconds. Creativity on Spielwerk will be endless with this
0:10

Okay this new model is insane. Generate an image, then give it to the model to turn it into a 3D model with matching texture in seconds. Creativity on Spielwerk will be endless with this

Eike Drescher

37,987 просмотров • 10 месяцев назад

When designing a room, choosing the right paint colours can completely transform how it feels & these 3 have work beautifully together to create a space that's rich, layered and timeless - I'm so pleased! I believe that soft earthy tones, with subtle...
0:10

When designing a room, choosing the right paint colours can completely transform how it feels & these 3 have work beautifully together to create a space that's rich, layered and timeless - I'm so pleased! I believe that soft earthy tones, with subtle...

Alice Gaskell •

16,641 просмотров • 8 месяцев назад

The new Krea 3D feature is insanely good. You can turn an image into a 3D model to move & rotate in space - and it guides real-time scene generation. Watch me move this couch around a living room ⬇️
0:18

The new Krea 3D feature is insanely good. You can turn an image into a 3D model to move & rotate in space - and it guides real-time scene generation. Watch me move this couch around a living room ⬇️

Justine Moore

32,737 просмотров • 1 год назад