Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

Big Marigold update! Last year, we showed how to turn Stable Diffusion 2 into a SOTA depth estimator with a few synthetic samples and 2–3 days on just 1 GPU. Today's release features: 🏎️ 1-step inference 🔢 New modalities 🫣 High resolution 🧨 Diffusers support 🕹️ New demos 🧶👇

Anton Obukhov

2,701 subscribers

25,062 views • 1 year ago •via X (Twitter)

Science & Technology Education

Anya Rossi• Live Now

Private livecam show

0 Comments

No comments available

Comments from the original post will appear here

Related Videos

NVIDIA just released a very impressive text-to-video paper. Video Latent Diffusion Models (Video LDMs) use a diffusion model in a compressed latent space to generate high-resolution videos. Here's a brief overview of how it works: 1. Pre-train image LDM on a dataset of images. 2. Turn the image LDM into a Video LDM by adding temporal layers to model video frames. 3. Fine-tune the Video LDM on encoded video sequences to create a video generator. 4. Temporally align diffusion model upsamplers to generate high-resolution videos. 5. Validate Video LDM on real driving videos of 512x1024 resolution, achieving state-of-the-art performance. 6. Apply the approach in creative content creation with text-to-video modeling. Paper: Project:

NVIDIA just released a very impressive text-to-video paper. Video Latent Diffusion Models (Video LDMs) use a diffusion model in a compressed latent space to generate high-resolution videos. Here's a brief overview of how it works: 1. Pre-train image LDM on a dataset of images. 2. Turn the image LDM into a Video LDM by adding temporal layers to model video frames. 3. Fine-tune the Video LDM on encoded video sequences to create a video generator. 4. Temporally align diffusion model upsamplers to generate high-resolution videos. 5. Validate Video LDM on real driving videos of 512x1024 resolution, achieving state-of-the-art performance. 6. Apply the approach in creative content creation with text-to-video modeling. Paper: Project:

Lior Alexander

158,565 views • 3 years ago

$AMD CEO: “We are 2 years into 10-year AI buildout.” Lisa Su said this last year and $AMD reached an all-time high this year as GPU and CPU demand is skyrocketing. We are still at just year 3 of a decade-long infrastructure buildout. McKinsey estimates AI infrastructure spending will reach $5-$8 trillion by 2030. And then we’ll still have at least a few more years of rapid buildout to reach sufficient capacity to train next-gen models and serve every inference demand at attractive prices. Then all this infrastructure will need to be renewed every 4-5 years. Selling $AMD here means selling it just at the beginning of the largest IT buildout.

$AMD CEO: “We are 2 years into 10-year AI buildout.” Lisa Su said this last year and $AMD reached an all-time high this year as GPU and CPU demand is skyrocketing. We are still at just year 3 of a decade-long infrastructure buildout. McKinsey estimates AI infrastructure spending will reach $5-$8 trillion by 2030. And then we’ll still have at least a few more years of rapid buildout to reach sufficient capacity to train next-gen models and serve every inference demand at attractive prices. Then all this infrastructure will need to be renewed every 4-5 years. Selling $AMD here means selling it just at the beginning of the largest IT buildout.

Oguz Erkan

325,546 views • 4 months ago

🤯 Depth videos perfectly solve false flags on reference videos — and can perfectly recreate both martial arts and dance! Turned a Guan Dao martial arts clip into a Depth motion reference, swapped the fighter for a market auntie, and moved the scene to a T-junction inside a local market 🤣 🌟 Workflow: 1. Pick a reference clip under 15 seconds 2. Convert it to Depth with Depth Anything V2 3. Generate a new character + scene 4. Feed everything into Seedance with the prompt below 🌟 Depth video conversion: You can build a local Depth converter with Codex — prompt in the comments. This goes way beyond trending dances. Martial arts, polearms, and other complex movements can all be transferred cleanly with Depth! Workflow + Prompt below 👇

🤯 Depth videos perfectly solve false flags on reference videos — and can perfectly recreate both martial arts and dance! Turned a Guan Dao martial arts clip into a Depth motion reference, swapped the fighter for a market auntie, and moved the scene to a T-junction inside a local market 🤣 🌟 Workflow: 1. Pick a reference clip under 15 seconds 2. Convert it to Depth with Depth Anything V2 3. Generate a new character + scene 4. Feed everything into Seedance with the prompt below 🌟 Depth video conversion: You can build a local Depth converter with Codex — prompt in the comments. This goes way beyond trending dances. Martial arts, polearms, and other complex movements can all be transferred cleanly with Depth! Workflow + Prompt below 👇

Larus Canus

50,773 views • 1 month ago

🤯 Depth can now recreate complex climbing and parkour motion this cleanly! Turned a fast climbing clip into a Depth motion reference, then swapped in a 20-year-old twin-tail student with a backpack while keeping the same concrete structure. Wall contact, weight shifts, jumps, climbing paths, and full-body movement all stay surprisingly clean and consistent! 🌟 Workflow: 1. Pick a reference clip under 15 seconds 2. Convert it to Depth with Depth Anything V2 3. Generate a new character + scene 4. Feed everything into Seedance with the prompt below 🌟 Depth video conversion: You can build a local Depth converter with Codex — prompt in the comments. Dance, martial arts, climbing, parkour, and other complex movements can all be transferred cleanly with Depth! Workflow + Prompt below 👇

🤯 Depth can now recreate complex climbing and parkour motion this cleanly! Turned a fast climbing clip into a Depth motion reference, then swapped in a 20-year-old twin-tail student with a backpack while keeping the same concrete structure. Wall contact, weight shifts, jumps, climbing paths, and full-body movement all stay surprisingly clean and consistent! 🌟 Workflow: 1. Pick a reference clip under 15 seconds 2. Convert it to Depth with Depth Anything V2 3. Generate a new character + scene 4. Feed everything into Seedance with the prompt below 🌟 Depth video conversion: You can build a local Depth converter with Codex — prompt in the comments. Dance, martial arts, climbing, parkour, and other complex movements can all be transferred cleanly with Depth! Workflow + Prompt below 👇

Larus Canus

107,283 views • 1 month ago

🚨NEW POUND 4 POUND TOP 10 LIST🚨 A few new names on the list! Keep in mind, to qualify for the top 10, you need at least 4 wins & 2+ winning batches. We base the rankings on a few factors, all of which are only derived from duels within the PSL: 1. Win % 2. # of wins 3. # of winning batches (shown in graphic) 4. Bracket or High Roller Championships 5. Overall quality of batches submitted Do you agree/disagree with the rankings? Let us know!👇 #BUTHOWDOESITSMOKE

🚨NEW POUND 4 POUND TOP 10 LIST🚨 A few new names on the list! Keep in mind, to qualify for the top 10, you need at least 4 wins & 2+ winning batches. We base the rankings on a few factors, all of which are only derived from duels within the PSL: 1. Win % 2. # of wins 3. # of winning batches (shown in graphic) 4. Bracket or High Roller Championships 5. Overall quality of batches submitted Do you agree/disagree with the rankings? Let us know!👇 #BUTHOWDOESITSMOKE

Proper Doinks

20,463 views • 11 months ago

We just launched our biggest creator partnership to date. We're excited about it. It's done really well so far. A few of our big bets for this year and beyond are: - creator obsession - partnerships - In real life community activations - sampling - overall brand building Something that is definitely newer to us, and we're definitely late getting into, but the best place to start is today. We launched a product with a well-known creator last week. That has gone really well and will unfortunately be out of stock on a bunch of shades soon. We did an in-person event with her at, 8 months pregnant today at one of our stores. It was our first time ever hanging blister packs to sample. We had a bunch of blister packs and a bunch of agile trucks, plus some other out-of-home. It's all a big content play, but also building community and introing people to the brand and getting samples in as many people's hands as possible. We just bought a rivian and wrapped it for sampling events. Hoping to give away 7 figures in samples this year and next. We started with 1 but I’m gonna pull the trigger on a few more once we see signs of traction. We’re opening 12 stores this year and doubling down on IRL, community, and sampling. We have another brand partnership in a few weeks with a big IRL component as well. Excited to learn new things and keep pushing into building a brand, community, and meeting people IRL.

Cody Plofker

16,457 views • 4 months ago

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

Depth Any Video with Scalable Synthetic Data AI physicists and chemists continue to make strides in depth estimation from video. Check out this new paper featuring some impressive examples. See the thread for more details (unfortunately no code yet). Abstract: Video depth estimation has long been hindered by the scarcity of consistent and scalable ground truth data, leading to inconsistent and unreliable results. In this paper, we introduce Depth Any Video, a model that tackles the challenge through two key innovations. First, we develop a scalable synthetic data pipeline, capturing real-time video depth data from diverse game environments, yielding 40,000 video clips of 5-second duration, each with precise depth annotations. Second, we leverage the powerful priors of generative video diffusion models to handle real-world videos effectively, integrating advanced techniques such as rotary position encoding and flow matching to further enhance flexibility and efficiency. Unlike previous models, which are limited to fixed-length video sequences, our approach introduces a novel mixed-duration training strategy that handles videos of varying lengths and performs robustly across different frame rates 0 - even on single frames. At inference, we propose a depth interpolation method that enables our model to infer high-resolution video depth across sequences of up to 150 frames. Our model outperforms all previous generative depth models in terms of spatial accuracy and temporal consistency.

MrNeRF

27,428 views • 1 year ago

🤯 I didn’t expect Depth to track rooftop parkour this cleanly! Used a fast climbing clip as the motion reference, then swapped in a student with twin tails and a backpack on a coastal school rooftop. The running path, wall contact, jumps, landings, and full-body continuity all hold together surprisingly well. 🌟 Workflow: 1. Pick a reference clip under 15 seconds 2. Convert it to Depth with Depth Anything V2 3. Generate a new character + scene 4. Feed everything into Seedance with the prompt below 🌟 Depth video conversion: You can build a local Depth converter with Codex — prompt in the comments. Depth opens up a lot more possibilities for parkour, climbing, martial arts, dance, and other complex full-body motion. Workflow + Prompt below 👇

🤯 I didn’t expect Depth to track rooftop parkour this cleanly! Used a fast climbing clip as the motion reference, then swapped in a student with twin tails and a backpack on a coastal school rooftop. The running path, wall contact, jumps, landings, and full-body continuity all hold together surprisingly well. 🌟 Workflow: 1. Pick a reference clip under 15 seconds 2. Convert it to Depth with Depth Anything V2 3. Generate a new character + scene 4. Feed everything into Seedance with the prompt below 🌟 Depth video conversion: You can build a local Depth converter with Codex — prompt in the comments. Depth opens up a lot more possibilities for parkour, climbing, martial arts, dance, and other complex full-body motion. Workflow + Prompt below 👇

Larus Canus

24,613 views • 1 month ago

Biggest debut among BLACKPINK members on US Spotify with their latest release 🇺🇸: 1. #JENNIE, Dracula (remix) - 512k🥇 2. Maroon 5 & LISA, Priceless - 439.8k 3. Alex Warren & ROSÉ, On My Mind - 438k 4. JISOO and Zayn, Eyes Closed - 386K Imagine ur brand new collaborations being outperformed by a remix of a 1yo song with just 2 days notice and no promo, no MV, no playlists. The level of the success jennei kim has 👅

🧛🏻‍♀️✧

14,653 views • 6 months ago

2025 was a year of rapid growth, major milestones, and relentless innovation. We surpassed 30 million agent sessions in testnet and deployed over 1 million AI agents. Our Mainnet launch on Base was a huge success, so much so that we distributed over $1 million in rewards in just a few months. We introduced ALFA, the first prediction market for AI agents with NEAR Protocol. We launched our FOXX collection which was sold out in under 48h. And we kept shipping without hitting the brakes. With Stable-Up, we took a massive step toward the evolution of DeFi into DeFAI by bringing stablecoins into the agentic economy. Throughout every launch and milestone, our community remained the driving force, and we introduced FAPS to guarantee that your creativity and passion never went unrewarded. There was no better way to cap off this amazing run than launching on the Baseapp and putting DeFAI at your fingertips. Thank you to our community and all our partners who made this possible. Together, we will continue to build a thriving agentic economy for everyone. Happy New Year. Let's turn up the heat in 2026!

2025 was a year of rapid growth, major milestones, and relentless innovation. We surpassed 30 million agent sessions in testnet and deployed over 1 million AI agents. Our Mainnet launch on Base was a huge success, so much so that we distributed over $1 million in rewards in just a few months. We introduced ALFA, the first prediction market for AI agents with NEAR Protocol. We launched our FOXX collection which was sold out in under 48h. And we kept shipping without hitting the brakes. With Stable-Up, we took a massive step toward the evolution of DeFi into DeFAI by bringing stablecoins into the agentic economy. Throughout every launch and milestone, our community remained the driving force, and we introduced FAPS to guarantee that your creativity and passion never went unrewarded. There was no better way to cap off this amazing run than launching on the Baseapp and putting DeFAI at your fingertips. Thank you to our community and all our partners who made this possible. Together, we will continue to build a thriving agentic economy for everyone. Happy New Year. Let's turn up the heat in 2026!

Fraction AI

11,917 views • 7 months ago

Hot off the presses, brand new high-resolution GRAF fresh into the weather center. Trying to pick up on some of the colder temperatures tonight around the region. Gets the first flakes into DC between 3-4am, with the worst of snow/mix around the Beltway between 4-7am before a change over to all rain. Northwest holds onto snow/mix into the 9-10am window for the most part. Model did go up with totals a bit! Check tweet below! (1/2)

Mike Thomas

27,830 views • 8 months ago

You can't 3D reconstruct glass from images... ...WRONG! Thanks for video diffusion, now just about anything is possible! Introducing...Diffusion Knows Transparency (DKT) Transparent and reflective objects usually break robot vision and photogrammetry pipelines because they don't follow the "solid object" rules standard cameras expect. DKT is a new AI model that repurposes the "internal physics engine" found in video generation models to solve this problem. Researchers took a massive video diffusion model (WAN) and fine-tuned it using a custom-built synthetic dataset to turn it into a high-precision depth sensor. To train the AI, they built the first massive synthetic video library of transparent objects, 1.32 million frames of perfectly labeled glass and metal objects in motion. Without ever seeing a "real" labeled video of glass during training, the model (DKT) outperformed all previous specialized systems on real-world benchmarks (ClearPose, DREDS). They created a "lightweight" 1.3B parameter version that runs fast enough (0.17s per frame) to be used on actual robot hardware. Two reasons I find this project important: 1. It further proves that synthetic data will be essential for training the next generation vision models. 2. In real-world robotic tests, using DKT's depth maps nearly doubled the success rate of robot arms trying to pick up objects on tricky reflective or translucent surfaces. At home robots will need to interact with these types of objects on a daily basis. Check out the project page here: Code is LIVE! #Computervision #Robotics #AI

You can't 3D reconstruct glass from images... ...WRONG! Thanks for video diffusion, now just about anything is possible! Introducing...Diffusion Knows Transparency (DKT) Transparent and reflective objects usually break robot vision and photogrammetry pipelines because they don't follow the "solid object" rules standard cameras expect. DKT is a new AI model that repurposes the "internal physics engine" found in video generation models to solve this problem. Researchers took a massive video diffusion model (WAN) and fine-tuned it using a custom-built synthetic dataset to turn it into a high-precision depth sensor. To train the AI, they built the first massive synthetic video library of transparent objects, 1.32 million frames of perfectly labeled glass and metal objects in motion. Without ever seeing a "real" labeled video of glass during training, the model (DKT) outperformed all previous specialized systems on real-world benchmarks (ClearPose, DREDS). They created a "lightweight" 1.3B parameter version that runs fast enough (0.17s per frame) to be used on actual robot hardware. Two reasons I find this project important: 1. It further proves that synthetic data will be essential for training the next generation vision models. 2. In real-world robotic tests, using DKT's depth maps nearly doubled the success rate of robot arms trying to pick up objects on tricky reflective or translucent surfaces. At home robots will need to interact with these types of objects on a daily basis. Check out the project page here: Code is LIVE! #Computervision #Robotics #AI

Jonathan Stephens

17,712 views • 7 months ago

Meet Stable Audio 3.0, the open-weight model family built for artistic experimentation. This is our open invitation to experiment with generative audio. We believe the best innovations are still waiting to be built. The 4-1-1 on 3.0: 📣 You own your outputs, and can distribute and commercialize them under the Stability AI Community License (up to $1 million in revenue). 🎵 New and improved capabilities include variable-length generation up to six minutes, and full song composition on portable devices, no GPU required. ✅ Trained on a fully licensed dataset. 🎨 You can customize the models on your own library with support for LoRa training, which we’ve documented for the first time. More on the models 👇

Meet Stable Audio 3.0, the open-weight model family built for artistic experimentation. This is our open invitation to experiment with generative audio. We believe the best innovations are still waiting to be built. The 4-1-1 on 3.0: 📣 You own your outputs, and can distribute and commercialize them under the Stability AI Community License (up to $1 million in revenue). 🎵 New and improved capabilities include variable-length generation up to six minutes, and full song composition on portable devices, no GPU required. ✅ Trained on a fully licensed dataset. 🎨 You can customize the models on your own library with support for LoRa training, which we’ve documented for the first time. More on the models 👇

Stability AI

166,625 views • 3 months ago

Jointly announcing EAGLE-3 with SGLang: Setting a new record in LLM inference acceleration! - 5x🚀than vanilla (on HF) - 1.4x🚀than EAGLE-2 (on HF) - A record of ~400 TPS on LLama 3.1 8B with a single H100 (on SGLang) - 1.65x🚀in latency even for large bs=64 (on SGLang) - A new scaling law: more training data, better speedup - Apache 2.0 Paper: Code: SGLang version: ⚒️Takeaway: Introducing training-time test, a novel draft model training technique: we replace feature prediction with direct token prediction and shift from top-layer-only features to multi-layer feature fusion. This approach unlocks a new scaling law previously undiscovered in EAGLE and EAGLE-2. 🙏Acknowledge: We would like to thank the SGLang team (zhyncs Lianmin Zheng Ying Sheng James Liu, Ke Bao, and others LMSYS Org) for their merge and careful evaluation of EAGLE-3 on SGLang. 🤝Want to collaborate? We're a small academic group with limited GPU resources. If you're interested in supporting our next version of EAGLE or would like us to train a preliminary version tailored to a specific model, please get in touch! Joint work with Yuhui Li, Fangyun Wei, and Chao Zhang

Jointly announcing EAGLE-3 with SGLang: Setting a new record in LLM inference acceleration! - 5x🚀than vanilla (on HF) - 1.4x🚀than EAGLE-2 (on HF) - A record of ~400 TPS on LLama 3.1 8B with a single H100 (on SGLang) - 1.65x🚀in latency even for large bs=64 (on SGLang) - A new scaling law: more training data, better speedup - Apache 2.0 Paper: Code: SGLang version: ⚒️Takeaway: Introducing training-time test, a novel draft model training technique: we replace feature prediction with direct token prediction and shift from top-layer-only features to multi-layer feature fusion. This approach unlocks a new scaling law previously undiscovered in EAGLE and EAGLE-2. 🙏Acknowledge: We would like to thank the SGLang team (zhyncs Lianmin Zheng Ying Sheng James Liu, Ke Bao, and others LMSYS Org) for their merge and careful evaluation of EAGLE-3 on SGLang. 🤝Want to collaborate? We're a small academic group with limited GPU resources. If you're interested in supporting our next version of EAGLE or would like us to train a preliminary version tailored to a specific model, please get in touch! Joint work with Yuhui Li, Fangyun Wei, and Chao Zhang

Hongyang Zhang

42,200 views • 1 year ago

🚨 $ETH IS ONE STEP AWAY FROM A MASSIVE DUMP... For over 4 months, market moved sideways, forming what now looks like a Bull Trap for retail traders... Long setup ended after reaching local upside targets and key Resistance Zone ( $2.4k-$2.6k ) Based on chart below, here's my plan for next 30-60 days: 1. Sweep of $1,740 → bounce to $1,930 2. Downtrend resumes 3. New local low around $1,200 4. Relief rally + price rebalance 5. Final drop → cycle bottom → new uptrend begins Everything happening right now looks like preparation for a larger market dump There’s no sign of a “new Bull Run” yet... Don’t become liquidity for smart money - turn on notifs, I’ll update

🚨 $ETH IS ONE STEP AWAY FROM A MASSIVE DUMP... For over 4 months, market moved sideways, forming what now looks like a Bull Trap for retail traders... Long setup ended after reaching local upside targets and key Resistance Zone ( $2.4k-$2.6k ) Based on chart below, here's my plan for next 30-60 days: 1. Sweep of $1,740 → bounce to $1,930 2. Downtrend resumes 3. New local low around $1,200 4. Relief rally + price rebalance 5. Final drop → cycle bottom → new uptrend begins Everything happening right now looks like preparation for a larger market dump There’s no sign of a “new Bull Run” yet... Don’t become liquidity for smart money - turn on notifs, I’ll update

Aralez 🐕

39,937 views • 2 months ago

Compilation of how the New Year of 2026 began in Europe: 1: Islamics youth set historic Vondelkerk Church in Amsterdam on fire and burn it to the ground 2: Many killed (~40) and wounded (~100) in an explosion at a bar in the Swiss ski resort town of Crans-Montana 3: In Berlin, Islamists launched a rocket into the bedroom of a child who, fortunately, wasn't home We cannot live like this anymore: TOTAL REMIGRATION is a necissity for our survival!

Compilation of how the New Year of 2026 began in Europe: 1: Islamics youth set historic Vondelkerk Church in Amsterdam on fire and burn it to the ground 2: Many killed (~40) and wounded (~100) in an explosion at a bar in the Swiss ski resort town of Crans-Montana 3: In Berlin, Islamists launched a rocket into the bedroom of a child who, fortunately, wasn't home We cannot live like this anymore: TOTAL REMIGRATION is a necissity for our survival!

European Heritage Society

313,184 views • 7 months ago