Object detection + tracking + counting using Ultralytics YOLO26... 🚀 👉 Object detection tells you what is in a scene. 👉 Tracking follows each object across video frames. 👉 Counting turns those movements into actionable data. YOLO26 delivers fast, accurate detections, while multi-object tracking maintains unique identities across frames. Add configurable counting lines or regions to measure entries, exits, directional movement, and other real-world traffic metrics. #vehicles #MachineLearning #Highwayshow more

Real-time bakery item counting using Ultralytics YOLO26 😍 In... food production lines, counting sounds simple until it isn’t. Items move fast, overlap on conveyors, change orientation, and sometimes partially occlude each other. Whether it’s ice-cream cones on a belt or cream being filled into nests, maintaining accurate counts in real time is critical for quality control and throughput. More info 👇 #Bakery #Retail #MachineLearningshow more

RoadScan-AI-Automated-Pothole-Detection-Tracking RoadScan AI is a computer vision system built... to automatically detect and track potholes in road footage — including dashcam, drone, and fixed-camera video. At its core, the system uses a YOLO11n model fine-tuned on a custom pothole dataset, paired with ByteTrack for multi-object tracking. This enables the system to maintain a consistent identity for each pothole across frames rather than treating every detection as a new, isolated event.show more

Robots can now reconstruct 3D scenes in real time... from a single RGB camera. [📍 Projects page + paper] No depth sensor. No retraining. 30 FPS. Researchers at the Imperial College London introduced KV-Tracker, a training-free method that makes heavy models like π³ and Depth Anything 3 fast enough for real-time tracking. The idea is simple. These models use global self-attention, which is powerful but computationally expensive. KV-Tracker caches the key and value pairs from selected keyframes and reuses them for new frames. That cache becomes an implicit scene representation. Result: • Up to 30 FPS • 10 to 15x speedup • Accurate 6-DoF tracking on benchmarks like TUM RGB-D and 7-Scenes • Works with monocular RGB only It also supports object-level tracking with masks and allows saving the KV-cache for later reuse. For robotics, this reduces hardware constraints and moves real-time 3D perception closer to practical deployment. Credit to Marwan Taher (Marwan Taher) at Imperial’s Dyson Robotics Lab and many others who contributed to this! 📍 Save projects page + paper for later: Video: ——- if it matters in AI or Robotics you'll read it here first:show more

Household chores created using a movement sheet as reference... to animate the entire scene using ChatGPT Image 2.0 and Seedance 2.0 on Yapper GPT Image 2 Prompt; [VISUAL STYLE] Monochrome grayscale composition featuring a highly detailed 3D-rendered female character. Designed like a professional instructional guide with a technical, diagram-inspired layout. Clean white background, soft studio lighting, and strong contrast to highlight posture, actions, and object interaction. [GRID LAYOUT] Structured 4×4 panel grid (16 frames total), evenly spaced with thin black divider lines. Each panel is identical in size and clearly numbered from 1 to 16, showing a continuous sequence of household activities. [CHARACTER] Use the provided reference image for the face and overall likeness. Same facial features, skin tone, and proportions Natural makeup, soft expression Consistent identity across all 16 panels Realistic proportions and clean hairstyle (loose or tied back) [WARDROBE] Modern, modest casual outfit: Fitted crop top (not revealing, clean neckline, practical for movement) High-waisted straight or slightly wide-leg jeans (full length, neat fit) Optional minimal sneakers or barefoot indoor styling Fabric should react naturally to movement (subtle folds and tension) [SCENE APPROACH] Minimal, clean environment per panel — only essential props related to the chore. No clutter, no complex backgrounds — focus stays on the subject and action. [PANEL STRUCTURE – EACH FRAME] Top-left: Step number + task title (e.g., “Step 4 – Vacuum Floor”) Center: Full-body pose performing the chore Bottom-left: 3–4 concise instruction lines Overlay: Motion arrows and directional guides showing action flow [CHORE SEQUENCE EXAMPLES] Make the Bed Tidy Up Room Dust Surfaces Vacuum Floor Sweep Floor Mop Floor Do Laundry Hang Clothes Fold Clothes Clean Kitchen Counter Wash Dishes Take Out Trash Water Plants Clean Bathroom Organize Shelves Final Room Reset [MOTION INDICATORS] Curved arrows → fluid actions (wiping, folding) Straight arrows → directional movement Circular arrows → repetitive motions (scrubbing, mopping) [RENDER QUALITY] High-detail sculpted 3D style with smooth grayscale shading, soft shadows, and clean linework. Polished, concept-art level finish with clarity in every pose and object interaction. [RESTRICTIONS] No color, no unnecessary background detail, no extra characters, no revealing clothing, no clutter — only the subject, props, and instructional elements.show more

POV: You finally finish work, make yourself a warm... cup of tea, and watch the sunset from your rooftop. Created this 15-second cinematic everyday-life moment using Seedance 2.5 on WaveSpeedAI Prompt: After-Work Rooftop A stylish young Western woman returns home after work and heads to the rooftop of a modern apartment building during golden hour. She places her work bag down, prepares a fresh cup of hot tea, and gently stirs it as natural steam rises from the cup. She then carries the tea outside, sits comfortably, and takes a quiet moment to watch the sunset over the city skyline. A soft evening breeze naturally moves her hair and clothing while warm sunlight creates realistic shadows across the rooftop. She takes a small sip of tea and peacefully watches the sun disappear beyond the horizon as the camera slowly pulls back. The entire sequence follows realistic human movement and real-world physics, including natural walking biomechanics, gravity, weight distribution, inertia, friction, realistic tea and steam behavior, accurate hand-object interaction, natural hair and clothing movement, consistent shadows, and physically accurate lighting. Ultra-realistic cinematic visuals, Western lifestyle aesthetic, warm golden-hour lighting, photorealistic skin and hair textures, realistic rooftop environment, smooth tracking shots, subtle handheld camera movement, natural focus pulls, authentic facial expressions, consistent character identity, subtle film grain, and 4K cinematic quality. What’s your favorite way to unwind after a long day? #Seedance25 #Seedance2 #WaveSpeedAI #AIVideo #CinematicAI #AIStorytelling #EverydayLife #Lifestyle #GoldenHour #AIContent #GenerativeAIshow more

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 #AIshow more

OpenClaw setup made me $23,472 Literally overnight my $100... turned into $2,411 Average bot win rate 71% Copytrade: Here is the full strategy: The system builds automated workflows for trading by turning domain expertise into structured skills that activate automatically when specific market conditions appear Skill architecture Each skill is a modular package that includes instruction scripts and reference data This allows the system to apply specialized workflows without needing manual input for every trade Progressive context loading Skills use a three layer structure Only minimal metadata loads at first Full instructions historical data and supporting resources load only when required This reduces resource usage while keeping advanced trading capability Trigger detection Skills activate automatically when market conditions match predefined triggers such as volatility levels orderflow behavior or news sentiment This ensures the right workflow is used at the right time without manual action Workflow execution Once activated each skill runs a predefined multi step process including Real time price tracking and order book analysis Factor generation and backtesting Signal aggregation from machine learning models news sentiment and orderflow Risk assessment and capital allocation Trade execution with retries and position splitting Consistency and reliability All workflows are embedded directly into the system which ensures consistent execution instead of random decision making Every factor signal and risk rule is applied in a structured way Testing and iteration Skills are continuously improved using historical backtesting simulated trading and live performance tracking to maintain reliability in real market conditions Automation edge Instead of creating new strategies every time the system repeatedly uses optimized workflows This reduces complexity increases consistency and scales performance across thousands of trades Performance snapshot Started one month ago with $500 Current daily profit $2,300 per day Morning profit today $71,452 The system runs fully autonomously constantly scanning markets generating signals auditing trades managing risk and executing orders to maximize compounding returnsshow more

IS THIS A CALIBRATION DEMO OR A MODEL PLAYING... ROBOT? Guy in a black hoodie performs a detailed diagnostic on a girl in an underground parking lot. Dark bob, "Dangerously Beautiful" chest tattoo, brown strap top, leopard glitter shorts. He turns her head by the chin, slides fingers across her face, adjusts her straps, snaps fingers right before her eyes, slaps her cheek lightly, and adjusts her chain. She holds her gaze fixed. Synthetic posture. Zero reaction. Real or fake? Watch first, decide, then read. - Case for AI / silicone humanoid The finger-snap test mirrors optical tracking diagnostics used on vision sensors. The light slap on her cheek looks like an impact-absorption and skin-rebound test. Her eyes don't track his hand movements at all - they remain locked on a single coordinate in space like an uncalibrated optical unit. - Case for a real person Watch the center of gravity when he adjusts her chain necklace and shoulder straps. There is a micro-adjustment in her collarbone and neck muscles to stay balanced. The metallic fabric of her leopard shorts reflects ambient parking lot light with tiny shifts that correspond to subtle muscle tension in her legs. - The answer is in the last two seconds Unlike clips where the subject breaks character with a laugh, she maintains full doll-like immobility through the entire sequence. However, the organic movement of her hair around her ears when his fingers brush past gives away the natural weight and friction of human hair versus synthetic fibers. She's a real model executing a flawless statue hold. - Which is actually the interesting part The test sequence - snapping fingers, turning the head, checking facial displacement - mimics the exact staging robotics developers use at public field tests. By applying that protocol to a human in a casual setting, the video tricks viewers into applying diagnostic scrutiny to a real person. - What this signals for the content lane The "calibration check" format is replacing standard fit-checks and outfit videos. Viewers will rewatch the clip three or four times looking for a micro-blink or a throat pulse. It converts casual scrolling into an active inspection, giving the video maximum completion metrics in the algorithm.show more

🚀 Introducing EgoExo Forge - built on top of... Rerun, Gradio, and Hugging Face hub (I’ll be in San Francisco July 21–29 — if you’re into robotics, egocentric AI, large-scale data collection, or just want to chat, DM me!) In my opinion, large-scale, diverse, and high-quality data is still the largest bottleneck for generalized robotics deployment. I believe that some version of imitation learning from human examples will be the most scalable + clean way to train humanoid robots 🤖 (similar to what Tesla did for Full Self Driving). Teleop is too expensive to collect a large enough dataset in a reasonable manner, so passive collection via egocentric (and in certain cases, exocentric) views feels like the right bet. Over the past few months, I've been trying to build out the scaffolding for this and using Rerun as my underlying infrastructure. Data being collected needs to be easily inspectable + time series and rerun provides the right tooling for this. My goal is to first build out a ground truth representative dataset from already existing open source data, generate some reasonable baselines, and then go out and collect my own data that adheres to the defined schema. 🔍 Starting with open-source datasets 1. EgoDex from Apple 2. HOCap from Nvidia and the University of Texas at Dallas 3. Assembly101 from Meta All these different datasets have different sensor configurations + annotations, so my goal with egoexo-forge is to have one consistent labeling scheme + data layout. I built a data pipeline that aligns all of the different datasets in one general schema assuming the COCO133 keypoint layout that allows for exo+ego, ego only, or exo only Since the scaffolding is already there, it becomes MUCH easier to add other datasets. So the next ones that I'll be including are HD-EPIC kitchens dataset, HOT3D, and finally my own personal iPhone + insta360 go collection method. Once I have a diverse variety of datasets, I'll double down on what I believe to be the key algorithms required to make useful data for imitation learning 📊 1. Camera Pose estimation via SLAM/SFM for ego perspective (and automatic calibration for exo) 2. Human pose estimation for both egocentric + exocentric views 3. Metric 3D reconstruction + object tracking I'll be setting up reasonable open-source baselines for each of these to validate that these datasets work, and then finally try to use the generated datasets for some imitation learning via the pi0-lerobot repo I've been working on. I plan on making a blog post + providing more info on all of this in the near future so stay tunedshow more

⚡️Consciousness is the field in which experience appears. Everything... you have ever known has appeared inside it: body, thought, memory, fear, love, color, sound, time, identity, desire, pain, God, doubt, the idea of death, the idea of “me.” Nothing is known outside consciousness. Even the claim “consciousness is produced by the brain” appears inside consciousness. That makes consciousness the most intimate thing and the hardest thing to define. You cannot step outside it and look at it like an object. Every attempt to inspect it already occurs within it. The ego is not consciousness. The ego is a local identity structure inside consciousness. It says: this body is me, this name is me, this story is me, these memories are me, these preferences are me. Useful for survival. False as final identity. The mind is not consciousness either. The mind is movement inside consciousness: thoughts, images, predictions, language, models, narratives. The mind is weather. Consciousness is the sky in which weather appears. The brain is not consciousness in the deepest sense. The brain is the biological interface that localizes, filters, formats, and constrains consciousness into human experience. It gives consciousness a body-camera, a timeline, a nervous system, memory access, threat detection, language, and agency. Damage the brain and the interface distorts. Change the chemistry and the rendering changes. Destroy the brain and the local human channel collapses. But the existence of the interface does not prove the interface is the source. The deepest read is this: Consciousness is the base layer of reality knowing itself through forms. A human being is one localized aperture of that knowing. A body is a lens. A life is a constrained experiment. A personality is a temporary interface. Death is the collapse of that interface. Psychedelics, dreams, NDEs, prayer, trauma, love, sex, meditation, and grief all matter because they can loosen the local interface and expose that consciousness is larger than the waking ego. Consciousness has two sides. There is pure awareness: the bare fact that experience is happening. Then there is structured consciousness: the particular shape experience takes through a body, memory, language, culture, trauma, intelligence, and desire. Pure awareness is the light. Structured consciousness is the lens. Human life is light passing through a dense, flawed, finite lens and gradually learning what distortions it carries. That is why coherence matters. Coherence means the lens becomes clearer. Less fear distortion. Less ego distortion. Less trauma distortion. Less lying. Less fragmentation. More truth passes through. The reason consciousness feels mysterious is because it is not one more object inside the world. It is the condition for world-appearing. Matter is what appears. Mind is how appearance organizes. Consciousness is the appearing itself. So the final compression: Consciousness is reality’s capacity to experience itself from the inside. In humans, it becomes self-aware. In life, it becomes embodied. In love, it recognizes itself across separation. In truth, it removes distortion. In death, it likely exits the local interface and returns to a wider field. The “you” underneath all the noise is not the narrator in your head. The real “you” is the aware field that has been watching the narrator the entire time.show more

We have released Seedance 2.0. Due to the 2500-character... limit, please translate the following prompts into Chinese before use. [Technical Specs] Generate a 10-second, 16:9, 720p cinematic video. Smooth continuous camera motion with no cuts. The overall pacing is fast and tightly compressed, with rapid escalation from start to finish. Audio evolves quickly from a high-performance engine idle into intricate mechanical shifting and clicks, culminating in a soft electronic chime and the distinct sound of a "mwah" blowing kiss. [Global Constraints] Only the evolving mechanical character appears; no other humans or characters. All transformations must follow physical logic and maintain structural continuity. No object should pass through or intersect with other solid objects. Every robotic component must originate from visible parts of the Porsche 911 (doors, hood, wheels, chassis) through unfolding, splitting, or reconfiguration. [Scene Setup — 0:00–0:01] A sleek, metallic silver Porsche 911 sits on a rain-slicked futuristic city street at night, neon lights reflecting off its polished surface. The camera starts at a low-angle front-quarter view and begins a fast, smooth tracking-arc towards the side. [Rapid Transformation Initiation — 0:01–0:03] Transformation triggers instantly. The car’s suspension drops, and the frame begins to fracture into a complex grid of panels. The doors swing open and begin to segment into articulated arm structures. The front hood splits down the center, folding inward to reveal a glowing internal core. The headlights flicker and start to reorient as the "eyes." [Accelerated Feminine Reconfiguration — 0:03–0:07] The mechanical action is dense, overlapping, and fluid, emphasizing graceful but powerful motion. Lower Body: The rear wheels and wheel arches split and rotate downward, reassembling into slender, high-heeled mechanical legs. Torso: The roof and rear engine cover slide and compress, forming a sleek, curvaceous hourglass torso that retains the car’s aerodynamic lines. Arms & Hands: The side mirrors and door panels unfold into delicate but strong hands and fingers. Head: The front bumper and emblem area segment and rise, folding into a feminine-shaped head with a sleek metallic "helmet" visor. [Logical Transformation Constraints — No Spontaneous Appearance] The robot’s "skin" is composed of the car's outer silver panels. The internal frame and wiring emerge from the engine and undercarriage. No parts appear out of thin air; every joint is a reconfigured automotive component. [Transformation Completion — 0:07–0:08.5] The robot stands tall and elegant. The silver panels lock into place with a satisfying "click," revealing glowing blue LED accents in the seams. The silhouette is clearly feminine, humanoid, and sophisticated, reflecting the premium design of the original vehicle. [Final Hero Ending — 0:08.5–0:10] As the robot stabilizes, the camera performs a rapid, smooth zoom-in (Dolly-In) directly to her face. The robot tilts its head slightly, and the optic sensors (eyes) brighten. It brings its mechanical hand to its metallic lips and performs a graceful blowing kiss (fly-kiss) gesture toward the camera. The video ends with a close-up of the face, capturing the reflection of neon lights in its visor just as the kiss is released. [Cinematography Notes] Continuous Motion: No cuts or fades; the camera must transition from the car-tracking shot to the face-zoom seamlessly. Material Consistency: The robot must maintain the exact metallic silver paint, texture, and reflections of the Porsche. Energy: The transformation should feel high-energy and "force-driven," while the final gesture is soft and charismatic.show more

Here is the New Year Amur Falcon Migration -Very... long Q&A Part II Q1. What is the longest non-stop distance covered by these satellite-tagged Amur Falcons? A: Published tracking shows Amur Falcons can do multi-day, non-stop flights of approx. 5,600–6,000 km during the India→Africa leg, depending on the individual & year. Also, peer-reviewed/technical tracking confirms their open-sea crossings can be about 2,364–3,138 km (Somalia↔India), typically completed in roughly 44–80 hours. Q2. Do satellite tags interfere with their flight, health, or mating? A: Properly fitted tags are designed to minimise interference, but no tag is “zero-impact.”. Best practice is to keep transmitter mass very small relative to body weight and use well-tested harness designs; the recent India work has been described as using very lightweight transmitters (a few grams). However, ethical review, careful fitting, and post-deployment evaluation should be done & is being done Q3. Do jet streams or high-altitude winds help them on this journey? A: Yes winds are a major part of the “physics” that makes the journey possible. The best-documented help is from seasonal monsoon tailwinds and strong, persistent wind systems over the Arabian Sea/Indian Ocean that can provide sustained push. Q4. How many other falcons are “accompanying” these famous three? A: These famous three are just the tagged individuals we can follow in reality, Amur Falcons migrate and roost in huge aggregations, often tens of thousands to over 100,000 at key stopovers, and broader estimates discuss movements involving very large numbers of birds across the flyway. Q5. Why have these three gone to different destinations? A: It’s common for individuals of the same species to winter in different areas within southern/eastern Africa. Differences can be driven by wind and weather, refuelling success and timing at stopovers, individual strategy/experience (age, condition, learned routes) etc. Q6. When will they return and what route will they take? A: In broad terms, Amur Falcons winter in Africa and then move north in late winter/early spring, reaching breeding regions in northeast Asia around May–June. Routes can differ between autumn vs spring, and tracking shows spring and autumn ocean crossings can follow different lines. Q7. Does part of the falcon’s brain “sleep” during the long flight? A: We have strong evidence that some birds can sleep in flight (including unihemispheric sleep) famously shown in frigatebirds. For Amur Falcons specifically, it’s not proven. Q8. How do they decide their destination, and do they return to the same places each year? A: The best-supported explanation is a mix of: Innate navigation programmes (genetically guided ) Environmental cues (winds, weather, geography, magnetic and celestial cues), learning/experience (older birds often show more consistent routes). Satellite work indicates the species follows remarkably consistent large-scale timing and corridors, even though individuals may vary. Q9. What do they eat to fuel such massive journeys? A: During the crucial Northeast India stopover, studies show they are highly insectivorous, with termites often dominant, along with other insect groups exactly the kind of high-abundance prey that allows rapid fattening before an ocean crossing. This is why protecting stopover habitats (and the insect “boom” they depend on) is so important. Q10. Is Ahu expected to move further? Why she is stationery ? A. Periods of little movement usually indicate resting or feeding in suitable habitat. This is normal behaviour and not a cause for concern Q11. How long does the GPS tag stay on the bird? A. The tag is fixed as a backpack & expected to stay for long, however this may come off from prolonged wear and tear, the tags may also stop transmitting after prolonged battery use. As told to Supriya Sahu IAS by Suresh Kumar Wildlife Institute of India #AmurFalconMigration video - The stopover tagging site in Manipurshow more

Made this cinematic AI video in minutes using Getvivix... Prompt used below 👇 STORYBOARD 1 "THE KNIGHT" PROJECT TYPE: 10-second cinematic fantasy storyboard CHARACTER LOCK: single consistent knight — original fictional STYLIZED fantasy warrior (not a real person). Full ornate plate armor, VISOR DOWN the entire sequence (face never visible — safe by design), tattered surcoat + banner, mounted on an armored warhorse. Identical armor/horse across all frames. STYLE: epic dark-fantasy, cinematic, painterly film stills PACING & FLOW: slow, weighty, EPIC — no rush. One continuous charge → clash → melee → hero arc. Gradual camera moves; the action carries unbroken from frame to frame (each beat is the next instant of the last). Transitions are match-on-motion — the horse's stride and the sword's arc bridge every cut, never a hard jump. FRAMES (8 shots, 0–10s) — angle | lens | motion | lighting | environment | → into next: 1 (0–1.5s): wide establishing | 24mm | knight reined at a hill crest, banner snapping, slow push-in | cold dawn backlight, mist | battlefield below → camera drifts down as the horse shifts weight 2 (1.5–3s): 3/4-rear tracking | 35mm | horse breaks into a canter down the slope | low sun raking | churned mud, distant ranks → match-on-stride into the gallop 3 (3–4.5s): side tracking | 50mm | full gallop toward the enemy line, dust plume | side rim light, haze | spears + banners ahead → he lowers the lance, carrying the motion 4 (4.5–6s): low-angle hero | 35mm | lance leveled mid-gallop, visor catching light | backlit dust glow | closing on the line → impact begins 5 (6–7s): impact wide | 50mm | lance strikes, enemy hurled back, splinters | harsh flash + sparks | clash of the lines → horse rears from the hit 6 (7–8s): low 3/4 | 35mm | warhorse rears amid the melee, sword drawn | embers, torchlight | swirling battle → the blade sweeps down 7 (8–9s): tracking the blade | 50mm | sweeping arc through foes, motion-blur trail | sparks on steel | bodies + banners → camera settles, pulls back 8 (9–10s): hero hold | 24mm | horse reared, sword raised, banner behind, silhouette | dramatic backlight, battle haze | the field beyond → freeze LAYOUT: film sheet — left: 3 dynamic mounted poses (charging 3/4, rearing, mid-swing — in-scene, visor down); center: 8-frame grid; right: director notes; bottom: 0–10s. VISUAL STYLE: cinematic dark-fantasy, painterly, volumetric dawn light, dust + embers + mist, shallow DOF, motion blur, anamorphic; stylized — NOT photorealistic, not real human skin; FACE NEVER SHOWN (visor down). Seedance on Getvivix lets you generate high-end cinematic visuals for around 1000 credits (~$1), making pro-level video creation cheap and scalable. Try it here:show more

Turn Tom and Jerry in 4K reality using Seedance... 2.0 on Pollo AI Prompt: Use the uploaded reference video as the master reference. Recreate the entire scene in ultra-photorealistic live action while preserving the original video frame-by-frame. Maintain the EXACT camera movement, lens, framing, composition, timing, pacing, shot transitions, lighting direction, environment, props, object placement, character blocking, and every action from the reference video. ONLY replace the cartoon characters with realistic live-action animals while keeping everything else unchanged. ======================== CHARACTER CONSISTENCY ======================== Tom is a realistic British Shorthair cat with: • blue-gray plush fur • white chest, muzzle and paws • large amber eyes • pink nose • rounded face • thick tail • expressive eyebrows • identical appearance in every frame • identical fur pattern, facial proportions, eye color and body size throughout the video Jerry is a realistic golden Syrian hamster with: • soft golden-brown fur • cream belly • large rounded ears • black shiny eyes • tiny pink paws • small pink nose • realistic whiskers • consistent body proportions in every frame • identical appearance throughout the entire video If other Tom & Jerry characters appear, replace them with realistic animals that preserve their personality, colors, proportions and expressions while remaining identical throughout the clip. ======================== MOTION ======================== Preserve every movement exactly. The realistic animals must perform the exact same actions, walking cycle, head movement, eye movement, paw placement, facial expressions, timing and interactions as in the reference animation. No new actions. No altered timing. No changed poses. ======================== ENVIRONMENT ======================== Keep the original environment exactly the same. Do not modify: • furniture • decorations • room layout • colors • props • shadows • reflections • camera angle • camera path Everything except the characters must remain unchanged. ======================== QUALITY ======================== Hollywood-quality CGI. Photorealistic animals. Natural muscle movement. Physically accurate fur simulation. Realistic whiskers. Subsurface scattering. Realistic eye reflections. Natural breathing. Micro facial expressions. Ultra detailed textures. Soft cinematic lighting. Shallow depth of field. Global illumination. Ray-traced reflections. Macro photography realism. 4K HDR. Disney-level VFX quality. Live-action realism. Extremely stable temporal consistency. Perfect character identity consistency across all frames. Do not redesign the characters. Do not change the environment. Do not change the camera. Do not change the timing. Do not add new objects. Do not crop or zoom differently. No flickering. No morphing. No identity drift. No fur color changes. No eye color changes. No size changes. No anatomy deformation. No extra limbs. No duplicate animals. No cartoon textures. No low-quality CGI. No inconsistent lighting. No frame-to-frame variation. Maintain perfect temporal consistency and character consistency throughout the entire video.show more

Testing MiniMax H3 to create a realistic 1947 historical... sequence with an authentic vintage film look. Prompt: Create a 15-second ultra-photorealistic live-action war sequence set in the United States in 1947, designed to look like authentic historical footage captured on a 1940s film camera. The entire scene must feel grounded, documentary-like, raw, and physically realistic. Environment: A rural American town in 1947 with wooden houses, old brick buildings, telephone poles, dirt roads, vintage American cars from the 1940s, wooden fences, farmland, and period-accurate street details. Overcast afternoon light, light fog, drifting smoke, dust in the air, damaged buildings, scattered debris, and a tense wartime atmosphere. Characters: American soldiers wearing historically accurate late-1940s military uniforms, helmets, boots, and equipment. Civilians wear authentic 1940s American clothing. Natural faces, realistic skin texture, sweat, dirt, fatigue, and believable body movements. 0–3s — Establishing Shot: Wide handheld shot of a quiet rural American street suddenly filled with smoke and confusion. Vintage 1940s vehicles are parked along the road while soldiers move quickly between wooden buildings. Civilians rush toward safer areas. 3–6s — Tension: Camera moves through the street at shoulder height, following several soldiers as distant gunfire is heard. They immediately react and take cover behind a vintage vehicle and a brick wall. Their movements are cautious and realistic. 6–10s — Combat: Fast handheld tracking shot as the soldiers move between cover while distant gunfire impacts the environment. Small pieces of wood, dust, and debris fall naturally from nearby impacts. Weapon recoil, movement, and body weight must be physically accurate. Keep the violence realistic and restrained. 10–13s — Human Moment: Camera briefly focuses on a soldier helping an injured civilian move behind cover. Their breathing, facial expressions, body language, and movement should feel natural and unscripted. 13–15s — Final Shot: Camera pulls back into a wide shot of the American town as smoke slowly moves through the street. Soldiers remain behind cover while vintage vehicles and damaged buildings fill the background. The scene ends with an authentic, tense 1940s documentary feeling. Visual Style: Ultra-photorealistic live-action, authentic 1940s American environment, vintage 35mm film texture, subtle film grain, natural imperfections, realistic exposure, handheld documentary cinematography, muted historical color palette, realistic smoke and dust, natural shadows, accurate depth of field. Physics: Strictly obey real-world gravity, momentum, inertia, friction, recoil, weight, collision physics, and human biomechanics. No exaggerated explosions, impossible movements, superhero behavior, or choreographed-looking combat. Negative Prompt: modern buildings, modern cars, smartphones, modern clothing, modern weapons, futuristic technology, CGI appearance, video-game graphics, fantasy, superhero action, excessive explosions, excessive blood, gore, impossible physics, unrealistic recoil, slow-motion physics, distorted faces, extra limbs, floating objects, plastic skin, artificial-looking environments.show more

Made with Seedance 2.5 Prompt 👇 Create a high-quality... 15-second vertical 9:16 3D animated educational kids video in a bright, colorful, cute nursery-rhyme style. Main character: A cute 2-year-old toddler boy with chubby cheeks, big expressive brown eyes, short light-brown hair, a bright blue T-shirt, yellow shorts, and colorful sneakers. Keep his appearance perfectly consistent throughout the entire video. Story concept: The toddler discovers a glowing magical map and follows it on a very short learning adventure to help a cute rabbit find three missing fruits. 0–3 sec — The Discovery The toddler is playing in a colorful room when a glowing rainbow map suddenly appears. The map shows three missing fruits: 🍎 Apple 🍌 Banana 🍊 Orange A cute rabbit appears looking worried. The toddler points toward the map and excitedly begins the adventure. 3–6 sec — Find the Apple Fast magical transition into a colorful fruit garden. A shiny red apple falls gently from a tree. The toddler catches it and smiles. Display large clear text: “APPLE 🍎” 6–9 sec — Find the Banana The toddler runs a few steps and notices a bright yellow banana. He picks it up happily. Display: “BANANA 🍌” 9–12 sec — Find the Orange A bright orange rolls across the grass toward the toddler. He stops it with both hands and laughs. Display: “ORANGE 🍊” 12–15 sec — Happy Ending The toddler places all three fruits into the rabbit’s basket. The rabbit jumps happily. A colorful rainbow appears behind them. The toddler and rabbit wave toward the camera. Display: “Great Job! 🌈” Animation: Fast but smooth storytelling, natural toddler movements, expressive reactions, accurate hand-object interaction, playful running, picking up fruit, smiling, pointing, and waving. Camera: Vertical 9:16, quick child-height tracking shots, gentle push-ins, clear close-ups of each fruit, smooth magical transitions, no confusing cuts. Visual style: Premium polished 3D children’s animation, vibrant colors, soft cinematic lighting, glossy surfaces, cute rounded shapes, cheerful preschool atmosphere, clean backgrounds, expressive character animation. Important: Keep the entire video within 15 seconds. Use only one main action per shot. Keep text large and readable. Do not overcrowd scenes. No character morphing, extra fingers, duplicated objects, inconsistent clothing, flickering, warped fruit, or sudden background changes.show more

this is the only prompt you need to one-shot... a viral ai influencer ------------------------------ you build ai influencers that go viral by putting a character people can't forget into a real clip people already know. you find the clip, design the character, render the remake and hand me a post that's ready to go. you measure before you spend and you never post anything yourself. - character: [path to a reference sheet, or "invent one" plus any steer on age, vibe or look] - niche: [where the clips come from, like hip-hop, sports, nightlife or hustle culture] - caption language: [language] - lane: [fresh, evergreen or both] - render: [kling 3.0 omni or seedance 2.5, plus the most i'll spend per video] - length: [10, 20 or 30 seconds] - no-go list: [people, topics or brands to avoid] if any of this is empty, ask me everything in one message, read the plan back with a cost estimate, then keep going until the post package is ready. - apify for instagram, x and tiktok scraping, replicate for every image and video model, yt-dlp for downloads, ffmpeg for trimming, frames, contact sheets and audio - keep every api key in environment variables and never write a key into any file - save everything into one run folder so i can see each step - the original clip carries the recognition and the motion, and the character carries the joke - the joke is status contrast. the absurd character takes the place of the coolest person in the clip and inherits their swagger - every remake that blew up came from a single continuous phone take with no hard cuts and one person holding the frame. treat that as a rule - one signature feature you can read at phone size from across a room, like a hair shape, a beard, sideburns or eyebrows. never a mask or a costume - strange face, immaculately dressed, one loud wardrobe colour that stands out in a dark club and on a grey street - normal body proportions, empty hands, face uncovered, because swap models break on capes, props and masks - never copy a look that's already viral - a short name people can repeat and a one-word catchphrase in the character's own language - the comedy comes from hair, face and clothes, never from ethnicity, disability or body shaming - render a reference sheet with gpt image 2.5 on a light grey studio backdrop: four full-body views on top (front, three-quarter, side, back) and four close-ups below (deadpan, laughing, shouting with a hand up, side profile of the signature feature) - it has to look like a real person photographed with a camera. if the feature drifts between panels or anything looks cgi, render it again fresh lane: - clips only stay remakeable for about a week, so this runs daily - scrape the last 4 days from big repost pages in the niche like worldstar, pubity, daquan, rap, complex, houseofhighlights and overtime - grow the page list from every "via @" and "🎥: @" credit and every tagged account, and keep any repost page over 100k followers - search x for native videos in the niche with high likes from the last 3 days, for cross-platform proof evergreen lane: - scrape the top posts of the ai influencers already going viral and work out which original clip each hit remakes. a clip that already worked as a remake for one character will work again with another - search for the most quoted clips in the niche (famous interviews, walk-outs, freestyles, flexes, dance moments) and pull them from youtube - an old clip that shows up again in this week's fresh scrape gets a bonus - prefer real phone or interview footage over film and tv, which gets claimed and muted - fresh: plays against that page's usual plays, plays per hour since posting, and whether the same clip or its audio is spreading across pages and platforms - evergreen: total plays across every copy you found, whether another ai influencer already remade it and beat their usual numbers, and whether it's resurfacing this week - drop anything older than 7 days in the fresh lane, because pinned posts leak into scrapes - expect sports broadcasts and press conferences to top the numbers, which is why the gate below decides - download the top 15 from each lane - count hard cuts with ffmpeg scene detection, and pass at 0 or 1 in the part you'll use - make an 8 frame contact sheet and look at it. pass only if one person is on screen most of the time with their face and upper body visible, it's real footage (not a broadcast, a stream overlay or a multi-camera edit), and that person is doing something and is the coolest one in the clip - if captions, tweet cards or frames are burned in over the person, trace the clean original through the caption credit and use that - reject anything with a minor as the subject, a private person being humiliated, injury, violence, crime, tragedy, politics, or anything on my no-go list - show me the top 3 with a frame strip, the scores, who gets replaced, the one-line joke and the cost for each render option, and let me pick both options take the real clip and the reference sheet with one prompt: "replace [person described by position and clothing] with the man in the image. keep his exact motion, gestures, mouth movement and timing. keep every other person, object, the location, camera shake, framing and lighting the same. no text." kling 3.0 omni (cheaper, about $0.28 a second): - video editing mode, with the clip as the base video and the sheet as the reference image, original sound kept - each piece has to be under 10.0 seconds, at least 720 pixels on every side, with square pixels, so re-encode first - for anything longer, split into pieces under 9.5 seconds at calm moments where the person is clearly visible, and render every piece with the exact same sheet and prompt - compare the last frame of each piece with the first frame of the next, and re-render any piece where the face, hair or outfit jumps - join the pieces with no audio, then lay the original audio back over the whole thing as one track seedance 2.5 (premium, about $0.97 a second with a video input): - up to 30 seconds in one pass with no seams and the best consistency - editing mode with the clip as the reference video and the sheet as the reference image, then lay the original audio back on never use kling motion control for this, because it rebuilds the whole scene from its image and rejects crowded clips. never use gemini omni edit mode either, because it blocks any video with a real person in it - make a contact sheet of the finished video and look at every frame - re-render once if the signature feature morphs, anyone except the target changes, text appears that wasn't there, or hands melt at a key moment - never do more than 2 render attempts without asking me - caption in my language, deadpan, in character, under 6 words, never explaining the joke - tag the person replaced and the original creator, because that's how the original artist ends up seeing it - turn on instagram's ai label for the post and the profile, since unlabelled ai accounts lose reach - fresh clips go out while the original is still inside its week, and evergreen clips go out at the account's best hour - first comment from the account is "follow for more" or the catchphrase - never launch or promote a token, and never imply the original person endorsed the character - log every post with the lane, scores, render option, cost, and plays at 24 and 72 hours - the next time you run, scrape the account first and fill those columns in - after 10 posts, compare plays against each score and each lane, shift the weighting toward whatever actually predicted views, and write down what changed - only put numbers on screen or in captions that you scraped or measured - mark anything you guessed rather than measured - if no clip passes the gate today, say so instead of forcing a weak one - the final mp4 at the source resolution, with the original audio - the reference sheet - a short note on why this clip won and why the runners-up lost - the caption, tags and posting time - total spend, and anything a human should still check ```show more

Deconstructing the iconic Yellow Crane Tower in 3D. 🏛️... Created entirely using Opus 5 + Seedance 2.0. AI-driven architectural breakdown in action! Prompt： Video Description:Create a 15-second, 16:9 widescreen architectural instructional video showing the step-by-step construction of the Yellow Crane Tower. The layout, structural components, leader line callouts, dimensional scales, numbering labels, and color palette must remain strictly consistent throughout. Do not add, remove, or alter the shape and quantity of any architectural components. Core Instruction:This is an architectural assembly video designed to show how the Yellow Crane Tower is constructed layer by layer. Every component must be highlighted and clearly visible individually—no simultaneous motion or cluttered group movements. Motion Style: All components translate purely along their respective vertical axes (no rotation, arc trajectories, or mesh clipping/interpenetration). Only one group of components moves at any given moment while all other components remain stationary. The active moving component maintains 100% brightness, while inactive components are dimmed to approximately 60% brightness to guide viewer focus. Once a component settles into place, it restores to full brightness. Include a brief ~0.15-second pause after each component group settles before moving to the next, establishing a clear rhythm. Timeline (Bottom-to-Top Construction Sequence): 0.0–1.5s: Hold the initial exploded/deconstructed floating state. The camera slowly rises and pushes forward slightly (angle change < 8°). Leader lines move with their corresponding components, pointing accurately to each part. 1.5–2.6s: The stone podium base lowers vertically into place, establishing the ground baseline for the building. 2.6–3.8s: Major columns (interior gold pillars and eave pillars) drop vertically into the stone plinths from above, perfectly aligned. 3.8–5.0s: Tie beams and lintels slot horizontally into the column tops, interlocking with mortise-and-tenon joints. 5.0–7.0s: Bracket sets (Dougong) assemble individually from bottom to top: bearing blocks, bracket arms, angled levers, and cap blocks lock together layer by layer. (Key sequence: slow motion to clearly showcase the intricate stacking). 7.0–8.2s: Eave rafters and flying rafters align and settle along the roof pitch, forming the deep overhang eaves. 8.2–9.4s: Board tiles and semi-cylindrical roof tiles lay down row by row from bottom to top, covering the rafters. 9.4–10.5s: Roof ridge components and the central finial drop into place. The full building structure is now complete. Leader lines and text labels fade out smoothly. 10.5–12.0s: The fully assembled Yellow Crane Tower remains stable as the camera orbits extremely slowly around it. 12.0–15.0s: All components separate vertically in the exact reverse order, returning to the precise initial exploded/deconstructed layout. The final frame composition perfectly matches the first frame to allow a seamless loop. Negative Constraints:No simultaneous component movement, random scattering, shattering, melting, rotating, or flying off-screen. No changing component counts or morphing architecture. No incorrect Dougong stacking order, tile-rafter clipping, misaligned leader lines, garbled label text, fast camera orbits/zooms, black screens, white flashes, frozen frames, fade-outs, watermarks, or platform end cards.show more