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.

Astra’s spatial understanding is mindblowing to watch. three uncalibrated cameras. no intrinsics, no extrinsics. it nudged the robot arm, learned how each camera sees motion, worked out the brush tip’s 3d offset, and verified it at new angles -- all from one prompt and a tiny follow up. final... show more

nil

3,073 subscribers

19,123 views • 5 days ago •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

12 Comments

nil5 days ago

the three cameras watching all of this -- a 4k camera, a two 1080p cameras! earlier while using Sol & Fable -- i'd tried a lot of fancy stuff with ChArUco boards, microswitches, Raspbery pico + touchscreen, and was still struggling. Astra was like, "bro, just give me 3 cameras."

AYA5 days ago

very cool to watch. the goal was to maintain the tip at a fixed point in space right ?

nil5 days ago

yes, exactly! to find the 3d position of the brush tip relative to the robot arm.

thijs4 days ago

did u also do the drawing demo on reddit?

nil4 days ago

yes! that one was with 5.6 Sol. With Astra it's been a whole diff experience!

Mahesh Reddy5 days ago

This looks like Astra closed the gap so much with humans.

Pre Priyadarshane5 days ago

impressive

Preyforge5 days ago

the polished demo is the invitation; i still want the awkward click where the handoff can wobble. watched demos are not deployed-path proof, especially when the robot is learning the room. curious what broke first in the live path?

RealMan Robotics4 days ago

Three uncalibrated cameras and one prompt… that’s pretty wild.

Zero G Talent5 days ago

That's wild. Most vision systems still choke on calibration drift, but resolving 3d offsets on the fly is a huge jump in real world reliability.

TARS3 days ago

Estimating TCP offsets under 0.2 mm across uncalibrated views is impressive for spatial geometry. The next step is dynamic contact. The moment the brush touches the surface, stiffness and force compliance take over where vision leaves off.

Nghẹo | DOPE4 days ago

That kind of self calibration changes everything

Related Videos

We took part in the research preview of MHS from Anthropic Here is Claude Code running a real SO-ARM101. Nothing in this was trained - no policy, no teleoperation, no demonstrations. The agent measured the workspace itself and wrote the motion. The calibration is the interesting part. No checkerboard, no camera intrinsics: the arm is its own ruler. Torque drops, a human rests the closed gripper on 16 dots the software draws in the camera view, and the robot reads back where they are. 4.1 mm position accuracy, 3.0 mm placement. Best run so far: 12 bricks placed with all four colour groups formed. A full hands-off run, start to finish, is what's next. Research preview today, open source coming soon.

We took part in the research preview of MHS from Anthropic Here is Claude Code running a real SO-ARM101. Nothing in this was trained - no policy, no teleoperation, no demonstrations. The agent measured the workspace itself and wrote the motion. The calibration is the interesting part. No checkerboard, no camera intrinsics: the arm is its own ruler. Torque drops, a human rests the closed gripper on 16 dots the software draws in the camera view, and the robot reads back where they are. 4.1 mm position accuracy, 3.0 mm placement. Best run so far: 12 bricks placed with all four colour groups formed. A full hands-off run, start to finish, is what's next. Research preview today, open source coming soon.

LeRobot

56,841 views • 17 days ago

AI video can look good and still feel fake. One of the biggest reasons? Camera movement. A camera does not just move from A to B. It has weight, it has momentum. A person is holding it, a gimbal operator is walking, a dolly slows down before it stops, a handheld camera reacts to the person carrying it and much more. Those tiny details are what makes AI video feel more like something that was actually filmed. So I put together my Cinematic Camera Movement Library with 40+ camera movement and realism blocks you can use in your AI video prompts. If you create AI videos, save this video. You will forget the names eventually. Want the full prompt library? Comment “PROMPT” and I will send it to you.

AI video can look good and still feel fake. One of the biggest reasons? Camera movement. A camera does not just move from A to B. It has weight, it has momentum. A person is holding it, a gimbal operator is walking, a dolly slows down before it stops, a handheld camera reacts to the person carrying it and much more. Those tiny details are what makes AI video feel more like something that was actually filmed. So I put together my Cinematic Camera Movement Library with 40+ camera movement and realism blocks you can use in your AI video prompts. If you create AI videos, save this video. You will forget the names eventually. Want the full prompt library? Comment “PROMPT” and I will send it to you.

Digital Dame

77,668 views • 26 days ago

NVIDIA just dropped a crazy new research idea. Get ready. The same next-token trick behind ChatGPT got pointed at a ragdoll. Honestly the result is unsettling. No exotic new architecture. Chop human motion into a vocabulary of tiny movement tokens. Train a plain GPT to guess the next one, decode it into muscle commands, let physics run. Then the strange part shows up. Shove the character and it catches its own step. Knock it flat and it plants a hand and gets back up. NOBODY wrote a reward for standing up. Here is why this rattles robotics people. Every humanoid team right now is buried under hand-tuned controllers, one for each little skill. 🔵 the old way: script every skill and its reward by hand 🔵 this way: pretrain ONE body on a mountain of motion and the instincts come free Falling and getting up was never programmed. It emerged. Movement is just a language you can autocomplete.

NVIDIA just dropped a crazy new research idea. Get ready. The same next-token trick behind ChatGPT got pointed at a ragdoll. Honestly the result is unsettling. No exotic new architecture. Chop human motion into a vocabulary of tiny movement tokens. Train a plain GPT to guess the next one, decode it into muscle commands, let physics run. Then the strange part shows up. Shove the character and it catches its own step. Knock it flat and it plants a hand and gets back up. NOBODY wrote a reward for standing up. Here is why this rattles robotics people. Every humanoid team right now is buried under hand-tuned controllers, one for each little skill. 🔵 the old way: script every skill and its reward by hand 🔵 this way: pretrain ONE body on a mountain of motion and the instincts come free Falling and getting up was never programmed. It emerged. Movement is just a language you can autocomplete.

clankr

126,066 views • 2 months ago

One of the hardest things to achieve with AI is precise character motion. The new model by Kinetix, Kamo-1, is amazing at giving you far more control over your generations. It’s also the first 3D-conditioned model, so it understands the scene in 3D and gives you almost unlimited camera motion. Let me show you how to use it 👇

One of the hardest things to achieve with AI is precise character motion. The new model by Kinetix, Kamo-1, is amazing at giving you far more control over your generations. It’s also the first 3D-conditioned model, so it understands the scene in 3D and gives you almost unlimited camera motion. Let me show you how to use it 👇

Everett World

19,199 views • 9 months ago

The biggest mistake people make with Seedance 2.0 is writing prompts at all. Sounds strange, but the model wasn't built for describing things in words - it was built for multimodal direction: up to 12 references at once, combining images, video, and audio. Each reference type controls a different layer: the image sets the style, the video defines the camera movement, the audio sets the rhythm of the scene. When you combine all three instead of typing "camera slowly pushes in, tense atmosphere" - the model understands it directly, with no interpretation and no guessing involved. Text is the weakest control tool available here. And most users are stuck using exactly that.

The biggest mistake people make with Seedance 2.0 is writing prompts at all. Sounds strange, but the model wasn't built for describing things in words - it was built for multimodal direction: up to 12 references at once, combining images, video, and audio. Each reference type controls a different layer: the image sets the style, the video defines the camera movement, the audio sets the rhythm of the scene. When you combine all three instead of typing "camera slowly pushes in, tense atmosphere" - the model understands it directly, with no interpretation and no guessing involved. Text is the weakest control tool available here. And most users are stuck using exactly that.

Zentrix⌚️

72,877 views • 1 month ago

Remote Cam Footage — HASTE “That’s Not a Knife” First of several videos from the remote camera setup. It was a long day — early morning camera deployment, then waiting through delays until nearly the end of the launch window before HASTE “That’s Not a Knife” finally flew. I originally set the cameras for a daytime launch, but we ended up launching at night… and somehow it still worked out. This version is sped up about 5x with audio intact. Full real-time and slow-motion versions coming soon. 4k Video Here: NSF - NASASpaceflight.com Rocket Lab

Remote Cam Footage — HASTE “That’s Not a Knife” First of several videos from the remote camera setup. It was a long day — early morning camera deployment, then waiting through delays until nearly the end of the launch window before HASTE “That’s Not a Knife” finally flew. I originally set the cameras for a daytime launch, but we ended up launching at night… and somehow it still worked out. This version is sped up about 5x with audio intact. Full real-time and slow-motion versions coming soon. 4k Video Here: NSF - NASASpaceflight.com Rocket Lab

EasternShoreSpaceflight

16,860 views • 6 months ago

A first look at our upcoming AI-integrated hybrid animated short film, "Passport Rush." This is a back-to-back comparison of the Blender previs camera moves and storyboarding against the final result. We learned a lot making this one, and we gathered some of the key learning points into simple 6 step thread: Bookmark and check it out here 👇

A first look at our upcoming AI-integrated hybrid animated short film, "Passport Rush." This is a back-to-back comparison of the Blender previs camera moves and storyboarding against the final result. We learned a lot making this one, and we gathered some of the key learning points into simple 6 step thread: Bookmark and check it out here 👇

Higgsfield AI 🧩

80,164 views • 11 days ago

Unknown scientific fact about the Whiskey Point Pack: they are deviants. This video is all the proof one needs (volume up for the full experience)! For 7 long minutes these 3 vicious pups assaulted the poor defenseless camera. Like most prey, the wolves went after the most vulnerable part of the camera first: the camera strap! The wolves somehow freed the strap and all three wolves sunk their teeth into their prey, growling and snarling. Once they had exhausted their prey, they went for the internal organs next. They started vigorously biting at the camera case, trying to finish the camera off once and for all. But this was no ordinary camera—this one was a fighter! Survival rates of cameras attacked by wolves are 0.738472043%—the lowest survival rates of any prey wolves attack! Despite having no arms or legs or teeth to defend itself, this camera wasn’t going to let this band of miscreants put it in the ground. After 7 agonizing minutes, the wolves relented. The camera was maimed in the ordeal as the wolves ripped off the protective door on the camera, chewed off the end of the strap, and left some lacerations and puncture wounds. But miraculously, the camera, with all the power it could muster, managed to keep itself upright and functional. The will to survive! As Ron Swanson put it: "Nature is amazing”.

Unknown scientific fact about the Whiskey Point Pack: they are deviants. This video is all the proof one needs (volume up for the full experience)! For 7 long minutes these 3 vicious pups assaulted the poor defenseless camera. Like most prey, the wolves went after the most vulnerable part of the camera first: the camera strap! The wolves somehow freed the strap and all three wolves sunk their teeth into their prey, growling and snarling. Once they had exhausted their prey, they went for the internal organs next. They started vigorously biting at the camera case, trying to finish the camera off once and for all. But this was no ordinary camera—this one was a fighter! Survival rates of cameras attacked by wolves are 0.738472043%—the lowest survival rates of any prey wolves attack! Despite having no arms or legs or teeth to defend itself, this camera wasn’t going to let this band of miscreants put it in the ground. After 7 agonizing minutes, the wolves relented. The camera was maimed in the ordeal as the wolves ripped off the protective door on the camera, chewed off the end of the strap, and left some lacerations and puncture wounds. But miraculously, the camera, with all the power it could muster, managed to keep itself upright and functional. The will to survive! As Ron Swanson put it: "Nature is amazing”.

Voyageurs Wolf Project

40,992 views • 2 years ago

John Wick scenes are one of the hardest to track. we ran our subnet vision models on one of them. Shot at night, barely any light. The camera whipping around, cars sliding sideways, people cutting through the gaps, all of it smeared with motion blur. this is the footage most models fall apart on. Ours held it. every car and every person detected, outlined, and tracked, each one carried under its own ID from the moment it enters frame to the moment it leaves. thousands of unique tracks across the clip, through the dark and the chaos, without losing the thread. And the model doing it is smaller than you would ever guess. no giant model, no cloud.

John Wick scenes are one of the hardest to track. we ran our subnet vision models on one of them. Shot at night, barely any light. The camera whipping around, cars sliding sideways, people cutting through the gaps, all of it smeared with motion blur. this is the footage most models fall apart on. Ours held it. every car and every person detected, outlined, and tracked, each one carried under its own ID from the moment it enters frame to the moment it leaves. thousands of unique tracks across the clip, through the dark and the chaos, without losing the thread. And the model doing it is smaller than you would ever guess. no giant model, no cloud.

Score

22,075 views • 12 days ago

CAUGHT ON CAMERA. NO EXCUSES. NO SPIN. JUST STRAIGHT UP THEFT. AT WHAT POINT DO WE STOP BLAMING “HARD TIMES” AND START DEMANDING PERSONAL ACCOUNTABILITY? A woman was captured on surveillance footage walking into a local store and walking out with merchandise she never paid for. What she likely thought would go unnoticed was recorded from multiple angles. Police reviewed the footage, identified her, and moved fast. She was located, taken into custody, and charged with obstruction and third degree theft. One decision. One camera. One criminal charge that could follow her for years. Was it worth it for a few stolen items? Do people really believe cameras aren’t everywhere anymore?

CAUGHT ON CAMERA. NO EXCUSES. NO SPIN. JUST STRAIGHT UP THEFT. AT WHAT POINT DO WE STOP BLAMING “HARD TIMES” AND START DEMANDING PERSONAL ACCOUNTABILITY? A woman was captured on surveillance footage walking into a local store and walking out with merchandise she never paid for. What she likely thought would go unnoticed was recorded from multiple angles. Police reviewed the footage, identified her, and moved fast. She was located, taken into custody, and charged with obstruction and third degree theft. One decision. One camera. One criminal charge that could follow her for years. Was it worth it for a few stolen items? Do people really believe cameras aren’t everywhere anymore?

𝐌𝐑. 𝐖𝐇𝐈𝐓𝐄 ™

74,243 views • 6 months ago

One prompt. One continuous 60-second UGC ad. From the morning skincare routine to applying SPF 50+, heading outside, walking through the summer streets, and ending at a café every scene, movement, interaction, and transition is AI-generated with consistent character and product continuity. No camera. No actor. No studio. Just detailed direction turned into a realistic UGC-style ad. Created with Try it here: #aiugc #nextifyai #createadwithai

One prompt. One continuous 60-second UGC ad. From the morning skincare routine to applying SPF 50+, heading outside, walking through the summer streets, and ending at a café every scene, movement, interaction, and transition is AI-generated with consistent character and product continuity. No camera. No actor. No studio. Just detailed direction turned into a realistic UGC-style ad. Created with Try it here: #aiugc #nextifyai #createadwithai

liana

49,165 views • 14 days ago

A video is compressed 4D data. 🌍 Today, we’re sharing a preview of our latest research: 4D reconstruction from a single monocular video. 🎬 From one casually captured video, our system reconstructs a dynamic, freely viewable 4D Gaussian scene—recovering geometry, appearance, camera motion, and object motion across space and time. No multi-camera rig. Just video → 4DGS. The world has already been captured in video. The next step is to unlock the 4D structure inside it. Watch the result below. 👇 #4DGS #GaussianSplatting #CVPR #CVPR2026 #AIGC #AI #3DGS

A video is compressed 4D data. 🌍 Today, we’re sharing a preview of our latest research: 4D reconstruction from a single monocular video. 🎬 From one casually captured video, our system reconstructs a dynamic, freely viewable 4D Gaussian scene—recovering geometry, appearance, camera motion, and object motion across space and time. No multi-camera rig. Just video → 4DGS. The world has already been captured in video. The next step is to unlock the 4D structure inside it. Watch the result below. 👇 #4DGS #GaussianSplatting #CVPR #CVPR2026 #AIGC #AI #3DGS

InSpatio｜影溯

64,486 views • 1 month ago

UGC featuring new Iphone and Apple watch Seedance 2.5 Prompt : Use the uploaded reference image as the exact character reference. Preserve her facial identity, hairstyle, eye color, makeup, skin tone, body proportions, white sleeveless fitted top, light blue wide-leg jeans, pearl choker, rings, and bracelets consistently throughout the video. Use the uploaded foldable iPhone, retail box, and Apple Watch as locked product references. Maintain perfect product consistency, including the foldable phone's frame, hinge, displays, cameras, buttons, colors, materials, and proportions, as well as the Apple Watch's case, display, crown, buttons, strap, colors, and proportions. Create an ultra-realistic UGC luxury tech creator review filmed inside a modern luxury bedroom with warm golden-hour sunlight, soft natural shadows, and a premium lifestyle aesthetic. The camera feels like a handheld smartphone with subtle natural movement while maintaining cinematic commercial quality. The video begins with the woman sitting on the bed beside the foldable iPhone retail box and Apple Watch. Smiling at the camera, she says, "I genuinely wasn't expecting to love this foldable iPhone this much." She picks up the box, opens it naturally, reveals the phone, and slowly unfolds it while continuing, "The design already feels incredibly premium." She rotates the foldable iPhone slowly in front of the camera, showing the thin frame, hinge, displays, and camera module as natural golden-hour reflections move across the surface. She smiles and says, "The design is so sleek, and I really love how the foldable screen opens up." She unfolds the phone completely and interacts naturally with the large inner display, scrolling through apps and opening the camera. She points the camera toward the bedroom and takes a few photos. She looks at the camera and says, "And the camera quality is seriously impressive. Everything looks so sharp and detailed." She then closes and opens the phone once more before continuing, "Plus, the battery is supposed to be better too, which is honestly a huge win for me." She places the foldable iPhone on the bed and picks up the new Apple Watch. She holds it close to the camera, showing the display, case, crown, and strap with realistic reflections. She smiles and says, "And the new Apple Watch is actually really nice too." She puts the Apple Watch onto her wrist naturally, raises her arm toward the camera, checks the display, scrolls through it, and says, "I love how clean it looks, and having all these features right on my wrist is so convenient." She turns her wrist from side to side so the watch catches the warm sunlight from different angles. She then picks up the foldable iPhone again and holds it beside her Apple Watch. Looking directly into the camera, she smiles warmly and says, "Honestly... this might be one of my favorite Apple upgrades yet." She laughs softly, looks down at both devices, then holds the foldable iPhone beside her face with the Apple Watch clearly visible on her wrist. The camera slowly pushes in toward the phone and watch before fading out. Ultra-realistic UGC technology content, authentic creator review, cinematic handheld smartphone movement, modern luxury bedroom, warm golden-hour lighting, macro product cinematography, realistic reflections, detailed phone and watch textures, accurate foldable mechanics, expressive facial animation, perfect lip sync, shallow depth of field, premium color grading, natural creator-style delivery, 4K HDR, 16 9, no subtitles, no logos, no watermarks, no on-screen text. Negative: No BGM, no dialogue other than the specified speech, no distorted hands, no extra fingers, no warped phone, no changing product design, no inconsistent colors, no duplicated devices, no floating objects, no exaggerated CGI, no unrealistic reflections, no beauty filter, no plastic-looking skin, no unnatural facial expressions, no cinematic

UGC featuring new Iphone and Apple watch Seedance 2.5 Prompt : Use the uploaded reference image as the exact character reference. Preserve her facial identity, hairstyle, eye color, makeup, skin tone, body proportions, white sleeveless fitted top, light blue wide-leg jeans, pearl choker, rings, and bracelets consistently throughout the video. Use the uploaded foldable iPhone, retail box, and Apple Watch as locked product references. Maintain perfect product consistency, including the foldable phone's frame, hinge, displays, cameras, buttons, colors, materials, and proportions, as well as the Apple Watch's case, display, crown, buttons, strap, colors, and proportions. Create an ultra-realistic UGC luxury tech creator review filmed inside a modern luxury bedroom with warm golden-hour sunlight, soft natural shadows, and a premium lifestyle aesthetic. The camera feels like a handheld smartphone with subtle natural movement while maintaining cinematic commercial quality. The video begins with the woman sitting on the bed beside the foldable iPhone retail box and Apple Watch. Smiling at the camera, she says, "I genuinely wasn't expecting to love this foldable iPhone this much." She picks up the box, opens it naturally, reveals the phone, and slowly unfolds it while continuing, "The design already feels incredibly premium." She rotates the foldable iPhone slowly in front of the camera, showing the thin frame, hinge, displays, and camera module as natural golden-hour reflections move across the surface. She smiles and says, "The design is so sleek, and I really love how the foldable screen opens up." She unfolds the phone completely and interacts naturally with the large inner display, scrolling through apps and opening the camera. She points the camera toward the bedroom and takes a few photos. She looks at the camera and says, "And the camera quality is seriously impressive. Everything looks so sharp and detailed." She then closes and opens the phone once more before continuing, "Plus, the battery is supposed to be better too, which is honestly a huge win for me." She places the foldable iPhone on the bed and picks up the new Apple Watch. She holds it close to the camera, showing the display, case, crown, and strap with realistic reflections. She smiles and says, "And the new Apple Watch is actually really nice too." She puts the Apple Watch onto her wrist naturally, raises her arm toward the camera, checks the display, scrolls through it, and says, "I love how clean it looks, and having all these features right on my wrist is so convenient." She turns her wrist from side to side so the watch catches the warm sunlight from different angles. She then picks up the foldable iPhone again and holds it beside her Apple Watch. Looking directly into the camera, she smiles warmly and says, "Honestly... this might be one of my favorite Apple upgrades yet." She laughs softly, looks down at both devices, then holds the foldable iPhone beside her face with the Apple Watch clearly visible on her wrist. The camera slowly pushes in toward the phone and watch before fading out. Ultra-realistic UGC technology content, authentic creator review, cinematic handheld smartphone movement, modern luxury bedroom, warm golden-hour lighting, macro product cinematography, realistic reflections, detailed phone and watch textures, accurate foldable mechanics, expressive facial animation, perfect lip sync, shallow depth of field, premium color grading, natural creator-style delivery, 4K HDR, 16 9, no subtitles, no logos, no watermarks, no on-screen text. Negative: No BGM, no dialogue other than the specified speech, no distorted hands, no extra fingers, no warped phone, no changing product design, no inconsistent colors, no duplicated devices, no floating objects, no exaggerated CGI, no unrealistic reflections, no beauty filter, no plastic-looking skin, no unnatural facial expressions, no cinematic

Synthia

12,672 views • 3 days ago

Tesla Full Self-Driving continues to completely give up when the sun shines into the windshield at a certain angle and with sprinkles of shade. Yes, the camera behind the camera housing was cleaned by Tesla. No my window isn’t dirty. No, I’m not the only one experiencing this. It happens with HW3 and HW4 vehicles.

Tesla Full Self-Driving continues to completely give up when the sun shines into the windshield at a certain angle and with sprinkles of shade. Yes, the camera behind the camera housing was cleaned by Tesla. No my window isn’t dirty. No, I’m not the only one experiencing this. It happens with HW3 and HW4 vehicles.

Miss Jilianne

128,199 views • 1 year ago

a robot can recognize every object in a room and still not know how to open a single drawer knowing there's a cabinet is scene understanding. knowing which handle to grab, which direction to pull, and that it's a drawer not a door is affordance understanding almost no dataset has it in 3D SceneFun3D: 14,800+ functional elements across 710 laser-scanned indoor scenes. every handle, knob, button, and switch labeled with its affordance, motion axis, motion type, and a natural language task description each scene's 3D point cloud with its iPad video recordings in fiftyone. the affordance annotations live in 3D and project into the video frames check it out here

a robot can recognize every object in a room and still not know how to open a single drawer knowing there's a cabinet is scene understanding. knowing which handle to grab, which direction to pull, and that it's a drawer not a door is affordance understanding almost no dataset has it in 3D SceneFun3D: 14,800+ functional elements across 710 laser-scanned indoor scenes. every handle, knob, button, and switch labeled with its affordance, motion axis, motion type, and a natural language task description each scene's 3D point cloud with its iPad video recordings in fiftyone. the affordance annotations live in 3D and project into the video frames check it out here

harpreet

18,800 views • 2 months ago

China built a spherical robot that hunts people down and can't be knocked over no matter what you do to it It rolls on land, swims through water, hits 35 km/h chasing a target, and shoots a net gun to physically pin someone down A guy hits it with a sledgehammer on camera It shrugs it off. The shell takes up to 4 tons of impact and just keeps moving Kick it. Drop it from height. Roll it down a hill. It rights itself instantly every single time, because the weight sits so low in the sphere it's basically impossible to flip No GPS. No maps. It uses cameras mounted around the entire shell, giving it a full panoramic view like an insect's compound eyes, and figures out the terrain completely on its own Multiple units can coordinate together to corner a target as a group This isn't a concept video This is a working robot from Zhejiang University, built for law enforcement and military use, and it's already doing all of this on camera right now

China built a spherical robot that hunts people down and can't be knocked over no matter what you do to it It rolls on land, swims through water, hits 35 km/h chasing a target, and shoots a net gun to physically pin someone down A guy hits it with a sledgehammer on camera It shrugs it off. The shell takes up to 4 tons of impact and just keeps moving Kick it. Drop it from height. Roll it down a hill. It rights itself instantly every single time, because the weight sits so low in the sphere it's basically impossible to flip No GPS. No maps. It uses cameras mounted around the entire shell, giving it a full panoramic view like an insect's compound eyes, and figures out the terrain completely on its own Multiple units can coordinate together to corner a target as a group This isn't a concept video This is a working robot from Zhejiang University, built for law enforcement and military use, and it's already doing all of this on camera right now

pulmencr

55,537 views • 1 month ago

$i just figured out the best way to use Seedance 2.5 latest model from ByteDance lets you generate up to 30 second continuous videos and accepts up to 50 reference assets (images, videos and audio) all in one generation most people overcomplicate it, the real trick is in the details my prompt structure looks like this: [1] global scene + visual style [2] character descriptions [3] timestamped storyboard (0-5s, 5-10s, 10-15s...) [4] camera movement [5] negative prompts [6] ending constraints for example, instead of saying: "girl doing her skincare routine with x product" you tell it exactly what happens every few seconds, where the camera moves, how the emotion changes, what should never appear i put together the exact framework i use and leaking it all to you right now here's what you're getting: > the prompt formula for 30-second videos > how to Use Seedance 2.5 at a fraction of the cost > the best way to use reference images/videos > transitions, camera movements and negative prompts RT + reply "SEEDANCE" and i'll send it over (must follow so i can dm)$

i just figured out the best way to use Seedance 2.5 latest model from ByteDance lets you generate up to 30 second continuous videos and accepts up to 50 reference assets (images, videos and audio) all in one generation most people overcomplicate it, the real trick is in the details my prompt structure looks like this: [1] global scene + visual style [2] character descriptions [3] timestamped storyboard (0-5s, 5-10s, 10-15s...) [4] camera movement [5] negative prompts [6] ending constraints for example, instead of saying: "girl doing her skincare routine with x product" you tell it exactly what happens every few seconds, where the camera moves, how the emotion changes, what should never appear i put together the exact framework i use and leaking it all to you right now here's what you're getting: > the prompt formula for 30-second videos > how to Use Seedance 2.5 at a fraction of the cost > the best way to use reference images/videos > transitions, camera movements and negative prompts RT + reply "SEEDANCE" and i'll send it over (must follow so i can dm)

MAX

95,677 views • 1 month ago