Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

🚀 Motion Control, Leveled Up Newly upgraded Motion Control is now live in Kling VIDEO 2.6! Experience precise, full control over every action & expression ✅ Full-Body Motions — Body movements captured in stunning detail ✅ Fast & Complex Actions — From martial arts to dances, nothing moves too... show more

Kling AI

134,605 subscribers

255,413 просмотров • 7 месяцев назад •via X (Twitter)

Игры Наука и технологии

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

Kling 2.6 Motion Control is out NOW 💥 Complex moves under control, with full-body sync and expression mapping Upload an image of your character and a motion reference to track face expressions, body motion, and hand moves. Up to 30-sec generations Available for all plans

Kling 2.6 Motion Control is out NOW 💥 Complex moves under control, with full-body sync and expression mapping Upload an image of your character and a motion reference to track face expressions, body motion, and hand moves. Up to 30-sec generations Available for all plans

Magnific

23,278 просмотров • 7 месяцев назад

Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation paper page: Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text, with methods that can generate character animations from short prompts and specified durations. However, using a single text prompt as input lacks the fine-grained control needed by animators, such as composing multiple actions and defining precise durations for parts of the motion. To address this, we introduce the new problem of timeline control for text-driven motion synthesis, which provides an intuitive, yet fine-grained, input interface for users. Instead of a single prompt, users can specify a multi-track timeline of multiple prompts organized in temporal intervals that may overlap. This enables specifying the exact timings of each action and composing multiple actions in sequence or at overlapping intervals. To generate composite animations from a multi-track timeline, we propose a new test-time denoising method. This method can be integrated with any pre-trained motion diffusion model to synthesize realistic motions that accurately reflect the timeline. At every step of denoising, our method processes each timeline interval (text prompt) individually, subsequently aggregating the predictions with consideration for the specific body parts engaged in each action. Experimental comparisons and ablations validate that our method produces realistic motions that respect the semantics and timing of given text prompts.

Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation paper page: Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text, with methods that can generate character animations from short prompts and specified durations. However, using a single text prompt as input lacks the fine-grained control needed by animators, such as composing multiple actions and defining precise durations for parts of the motion. To address this, we introduce the new problem of timeline control for text-driven motion synthesis, which provides an intuitive, yet fine-grained, input interface for users. Instead of a single prompt, users can specify a multi-track timeline of multiple prompts organized in temporal intervals that may overlap. This enables specifying the exact timings of each action and composing multiple actions in sequence or at overlapping intervals. To generate composite animations from a multi-track timeline, we propose a new test-time denoising method. This method can be integrated with any pre-trained motion diffusion model to synthesize realistic motions that accurately reflect the timeline. At every step of denoising, our method processes each timeline interval (text prompt) individually, subsequently aggregating the predictions with consideration for the specific body parts engaged in each action. Experimental comparisons and ablations validate that our method produces realistic motions that respect the semantics and timing of given text prompts.

AK

126,595 просмотров • 2 лет назад

This work is not about a new technique. GMT (General Motion Tracking) shows good engineering practices that you can actually train a single unified whole-body control policy for all agile motion, and it works in the real world, directly with sim2real without adaptation. This is different from many existing works that trains/adapts one individual policy for each specific motion. Instead of just showing the beautiful motions/dances, this is actually practical. Because now you can convert all types of commands into real actions in real time without tuning anything. Visit to check out how we handle the following challenges: ✅ Partial Observability. ✅ Hardware Limitations. ✅ Unbalanced Data Distribution. ✅ Model Expressiveness. Work led by Zixuan Chen Mazeyu Ji

This work is not about a new technique. GMT (General Motion Tracking) shows good engineering practices that you can actually train a single unified whole-body control policy for all agile motion, and it works in the real world, directly with sim2real without adaptation. This is different from many existing works that trains/adapts one individual policy for each specific motion. Instead of just showing the beautiful motions/dances, this is actually practical. Because now you can convert all types of commands into real actions in real time without tuning anything. Visit to check out how we handle the following challenges: ✅ Partial Observability. ✅ Hardware Limitations. ✅ Unbalanced Data Distribution. ✅ Model Expressiveness. Work led by Zixuan Chen Mazeyu Ji

Xiaolong Wang

83,261 просмотров • 1 год назад

Humanoids finally move like humans… and can do more than copy. [Details + demos in thread 👇] A new framework, BeyondMimic, shows how to learn naturalistic whole-body control from human motion. But then goes further by composing those skills into versatile, zero-shot policies. Two key pieces make it work: •Motion tracking pipeline: robustly turns human mocap into dynamic robot motions (jump spins, sprinting, cartwheels) on real hardware •Guided diffusion policy: distills motion primitives and enables zero-shot downstream tasks with simple cost functions The result: ✅ State-of-the-art motion quality from mocap tracking ✅ Versatile task execution; waypoint navigation, joystick teleop, obstacle avoidance ✅ Stable sim-to-real transfer for diverse humanoid skills Thank you for sharing, Qiayuan Liao! 📍 Paper: Video:

Humanoids finally move like humans… and can do more than copy. [Details + demos in thread 👇] A new framework, BeyondMimic, shows how to learn naturalistic whole-body control from human motion. But then goes further by composing those skills into versatile, zero-shot policies. Two key pieces make it work: •Motion tracking pipeline: robustly turns human mocap into dynamic robot motions (jump spins, sprinting, cartwheels) on real hardware •Guided diffusion policy: distills motion primitives and enables zero-shot downstream tasks with simple cost functions The result: ✅ State-of-the-art motion quality from mocap tracking ✅ Versatile task execution; waypoint navigation, joystick teleop, obstacle avoidance ✅ Stable sim-to-real transfer for diverse humanoid skills Thank you for sharing, Qiayuan Liao! 📍 Paper: Video:

Ilir Aliu - eu/acc

72,522 просмотров • 1 год назад

This Google Veo 3 "frames to video" feature is pure genius 💡 If you're creating video ads, this Veo 3 technique gives you surgical control over every scene element. This frames-to-video feature will take any static product shot... And turn it into cinematic reels with precise camera movements & timing. Perfect for creative strategists, agencies, and DTC operators running video campaigns. But instead of hoping AI generates what you hope for... Just upload your reference image and get exact scene control in minutes. The game-changer? Your static assets become dynamic video gold → organized by camera moves, action sequences, & precise timing control. Here's what frames-to-video gives you: → upload any product reference image → add numbered annotations for camera movements → structure scene directions (I use JSON for precision) → set duration, fps, lighting specs → get frame-perfect video execution → no video shooting or expensive production needed Now you've got director-level control from single images. These HexClad reels were generated from two static product shots + a JSON prompt for each. When your campaign needs cinematic quality but you only have product photos? This frames-to-video workflow changes everything. And it's only getting better Want my full annotation setup guide, including the JSON prompts? Comment "FRAMES" + follow and I'll send it over (must be following/connected so I can DM)

This Google Veo 3 "frames to video" feature is pure genius 💡 If you're creating video ads, this Veo 3 technique gives you surgical control over every scene element. This frames-to-video feature will take any static product shot... And turn it into cinematic reels with precise camera movements & timing. Perfect for creative strategists, agencies, and DTC operators running video campaigns. But instead of hoping AI generates what you hope for... Just upload your reference image and get exact scene control in minutes. The game-changer? Your static assets become dynamic video gold → organized by camera moves, action sequences, & precise timing control. Here's what frames-to-video gives you: → upload any product reference image → add numbered annotations for camera movements → structure scene directions (I use JSON for precision) → set duration, fps, lighting specs → get frame-perfect video execution → no video shooting or expensive production needed Now you've got director-level control from single images. These HexClad reels were generated from two static product shots + a JSON prompt for each. When your campaign needs cinematic quality but you only have product photos? This frames-to-video workflow changes everything. And it's only getting better Want my full annotation setup guide, including the JSON prompts? Comment "FRAMES" + follow and I'll send it over (must be following/connected so I can DM)

Mike Futia

39,838 просмотров • 1 год назад

🔴 INTRODUCING BLOOD v3 🔴 The most advanced Solana bot just got even better. After months of relentless work, testing, and fine-tuning, we're proud to unveil BLOOD GUI – the most intuitive, lightning-fast, and powerful UI ever created. Built for speed, simplicity, and precision, it redefines the way you interact with bot. Newcomers can set everything up in minutes, while pros will execute their strategies with unmatched efficiency. 🔥 WHAT’S NEW? 🔥 ✅ Informative homepage – get all essential bot details at a glance. ✅ Sleek, optimized UI – clean design with zero performance loss. ✅ Smart settings management – hide unnecessary options or find the right setting instantly with a search bar. ✅ Enhanced position control – manage multiple trading modes with full control over your trades. ✅ 🔁 Unlimited wallet copying – mirror an infinite number of wallets. ✅ ⚡️ Upgraded nodes – FREE NODE + FASTEST GEYSERS FOR ALL HOLDERS. ✅ 📊 PNL & Winrate analytics – track performance, optimize strategies, and maximize profits with in-depth wallet analysis. And the best part? You get ALL of this with ZERO TRADING TAXES 💰. 🚨 This is not just an update. It’s a GAME-CHANGER. 🚨 There is also one more important thing…

🔴 INTRODUCING BLOOD v3 🔴 The most advanced Solana bot just got even better. After months of relentless work, testing, and fine-tuning, we're proud to unveil BLOOD GUI – the most intuitive, lightning-fast, and powerful UI ever created. Built for speed, simplicity, and precision, it redefines the way you interact with bot. Newcomers can set everything up in minutes, while pros will execute their strategies with unmatched efficiency. 🔥 WHAT’S NEW? 🔥 ✅ Informative homepage – get all essential bot details at a glance. ✅ Sleek, optimized UI – clean design with zero performance loss. ✅ Smart settings management – hide unnecessary options or find the right setting instantly with a search bar. ✅ Enhanced position control – manage multiple trading modes with full control over your trades. ✅ 🔁 Unlimited wallet copying – mirror an infinite number of wallets. ✅ ⚡️ Upgraded nodes – FREE NODE + FASTEST GEYSERS FOR ALL HOLDERS. ✅ 📊 PNL & Winrate analytics – track performance, optimize strategies, and maximize profits with in-depth wallet analysis. And the best part? You get ALL of this with ZERO TRADING TAXES 💰. 🚨 This is not just an update. It’s a GAME-CHANGER. 🚨 There is also one more important thing…

BLOOD SOLUTIONS

12,222 просмотров • 1 год назад

A team tested Pi0, Pi0 Fast, Gr00t, and ACT on real robot arms in manufacturing tasks. (🔖 Bookmark this for later!) The task was precise: place thin rectangular frames from a messy stack into a holder. The team fine-tuned each model on 100 real trajectories and compared training time, inference speed, motion quality, and success rates. ⬇️ Here’s a breakdown of what they found Pi0 (Original) ✅ Strongest overall performance in precise pick-and-place ✅ High success rate even in edge cases ✅ Longest training time (~11 hours, ~$30 per run) ✅ Inference time of 80 ms causes short pauses between actions Despite delays, it handles complex scenarios well… solid for high-precision tasks, but slow to train. Gr00t ✅ Trains fast (~2 hours, ~$5 per run) ✅ Performs almost as well as Pi0 on large-object tasks ✅ Struggles with fine precision; random movement in some trials ✅ More training didn’t fix jitter or random offsets Best suited for tasks where exact precision isn’t critical. Not ready for manufacturing-grade accuracy without more tuning. Pi0 Fast ✅ Promised faster training, but results were underwhelming ✅ Training at 6 hours still showed low success rates ✅ Inference was slower than expected ✅ Not reliable for generalizing even slightly new tasks Currently too unstable for real-world deployment. Doesn’t live up to the “Fast” name yet. ACT (Baseline) ✅ 200MB model—lightweight, but limited ✅ Struggles with stacked objects or ambiguous scenes ✅ Success rates around 70% in best-case setups ✅ Can’t match newer models on precision or generalization Still a solid baseline, but clearly a generation behind in robustness. 🚨 Extra Notes All newer models share a common issue: •Inference takes longer than a frame (80 ms vs 33 ms), so robots “pause” between chunks. •This results in jittery movements, but not a dealbreaker unless tasks are time-sensitive. Language-conditioned tasks also fell short: after training on two labeled tasks, the model couldn’t generalize to a third unseen combination using only text prompts. ✅ The good news? These models adapt well to new robot arms with quick fine-tuning. ❌ The bad news? There’s still no plug-and-play solution for improving performance after deployment. Reinforcement learning or DAgger-style data collection during real-world operation may be the next big step, something many teams in robotics are actively working on.

A team tested Pi0, Pi0 Fast, Gr00t, and ACT on real robot arms in manufacturing tasks. (🔖 Bookmark this for later!) The task was precise: place thin rectangular frames from a messy stack into a holder. The team fine-tuned each model on 100 real trajectories and compared training time, inference speed, motion quality, and success rates. ⬇️ Here’s a breakdown of what they found Pi0 (Original) ✅ Strongest overall performance in precise pick-and-place ✅ High success rate even in edge cases ✅ Longest training time (~11 hours, ~$30 per run) ✅ Inference time of 80 ms causes short pauses between actions Despite delays, it handles complex scenarios well… solid for high-precision tasks, but slow to train. Gr00t ✅ Trains fast (~2 hours, ~$5 per run) ✅ Performs almost as well as Pi0 on large-object tasks ✅ Struggles with fine precision; random movement in some trials ✅ More training didn’t fix jitter or random offsets Best suited for tasks where exact precision isn’t critical. Not ready for manufacturing-grade accuracy without more tuning. Pi0 Fast ✅ Promised faster training, but results were underwhelming ✅ Training at 6 hours still showed low success rates ✅ Inference was slower than expected ✅ Not reliable for generalizing even slightly new tasks Currently too unstable for real-world deployment. Doesn’t live up to the “Fast” name yet. ACT (Baseline) ✅ 200MB model—lightweight, but limited ✅ Struggles with stacked objects or ambiguous scenes ✅ Success rates around 70% in best-case setups ✅ Can’t match newer models on precision or generalization Still a solid baseline, but clearly a generation behind in robustness. 🚨 Extra Notes All newer models share a common issue: •Inference takes longer than a frame (80 ms vs 33 ms), so robots “pause” between chunks. •This results in jittery movements, but not a dealbreaker unless tasks are time-sensitive. Language-conditioned tasks also fell short: after training on two labeled tasks, the model couldn’t generalize to a third unseen combination using only text prompts. ✅ The good news? These models adapt well to new robot arms with quick fine-tuning. ❌ The bad news? There’s still no plug-and-play solution for improving performance after deployment. Reinforcement learning or DAgger-style data collection during real-world operation may be the next big step, something many teams in robotics are actively working on.

Ilir Aliu

21,844 просмотров • 1 год назад

🎬 Global Launch: Dreamina Seedance 2.5 is now live! Our most cinematic video model yet has arrived, and Dreamina AI is your official platform. Step into a new era of precision control and long-take storytelling: · Native 30s videos in a single generation · A new interactive editing experience: reshape any part of the frame · Long video mode: up to 3 minutes, with consistency that holds throughout · Dreamina AI plugins for Maya and Blender, built for film-grade production pipelines · Up to 50 multimodal references, now with support for 3D white models and green-screen footage — bringing AI generation into real production workflows · More true-to-life lighting and shadows, richer motion detail, and precise timestamp control down to 1 second Dreamina Seedance 2.5 is more than a model. It's an upgrade to what you can create. Now open to Dreamina AI subscribers. 🕐 Available now across Southeast Asia, the Middle East, Africa, Europe, and South America. More countries and regions coming soon. ✨ Watch Aura Design in action with the new model ⏳ Limited time: save up to 32% on credits per generation 🎁 Repost and comment – first 300 users get 1,000 credits in 12 hours #DreaminaSeedance25 #DreaminaAI

🎬 Global Launch: Dreamina Seedance 2.5 is now live! Our most cinematic video model yet has arrived, and Dreamina AI is your official platform. Step into a new era of precision control and long-take storytelling: · Native 30s videos in a single generation · A new interactive editing experience: reshape any part of the frame · Long video mode: up to 3 minutes, with consistency that holds throughout · Dreamina AI plugins for Maya and Blender, built for film-grade production pipelines · Up to 50 multimodal references, now with support for 3D white models and green-screen footage — bringing AI generation into real production workflows · More true-to-life lighting and shadows, richer motion detail, and precise timestamp control down to 1 second Dreamina Seedance 2.5 is more than a model. It's an upgrade to what you can create. Now open to Dreamina AI subscribers. 🕐 Available now across Southeast Asia, the Middle East, Africa, Europe, and South America. More countries and regions coming soon. ✨ Watch Aura Design in action with the new model ⏳ Limited time: save up to 32% on credits per generation 🎁 Repost and comment – first 300 users get 1,000 credits in 12 hours #DreaminaSeedance25 #DreaminaAI

Dreamina AI

595,252 просмотров • 13 дней назад

🚀 Introducing HunyuanCustom: An open-source, multimodal-driven architecture for customized video generation, powered by HunyuanVideo-13B. Outperforming existing open-source models, it rivals top closed-source solutions! 🎥 Highlights: ✅Subject Consistency: Maintains identity across single & multi-subject video generation. ✅Multimodal Inputs: Supports text, images, audio, and video for highly controlled outputs. ✅High-Quality Output: Industry-leading detail, motion smoothness, and lighting realism. Key Features: 1️⃣Single-Subject Video: Upload an image + text (e.g., “He’s walking a dog”) to create coherent videos with new actions, outfits, and scenes. 2️⃣Multi-Subject Video: Generate videos with multiple subjects (e.g., a man drinking coffee in a cozy room) from separate image inputs. 3️⃣Audio-Driven Video: Sync audio with visuals for talking or singing in any scene—perfect for digital avatars, virtual customer service, and more. 4️⃣Video-Driven Video: Seamlessly insert or replace subjects into any video for creative enhancements. Dive Deeper: 🌐Project: 📝Technical Details: 💻Code: The single-subject video capability is open-sourced and live on the Hunyuan website, with more features to be released soon! 👉 Try it now: #HunyuanCustom #VideoGeneration #AI #TencentHunyuan

🚀 Introducing HunyuanCustom: An open-source, multimodal-driven architecture for customized video generation, powered by HunyuanVideo-13B. Outperforming existing open-source models, it rivals top closed-source solutions! 🎥 Highlights: ✅Subject Consistency: Maintains identity across single & multi-subject video generation. ✅Multimodal Inputs: Supports text, images, audio, and video for highly controlled outputs. ✅High-Quality Output: Industry-leading detail, motion smoothness, and lighting realism. Key Features: 1️⃣Single-Subject Video: Upload an image + text (e.g., “He’s walking a dog”) to create coherent videos with new actions, outfits, and scenes. 2️⃣Multi-Subject Video: Generate videos with multiple subjects (e.g., a man drinking coffee in a cozy room) from separate image inputs. 3️⃣Audio-Driven Video: Sync audio with visuals for talking or singing in any scene—perfect for digital avatars, virtual customer service, and more. 4️⃣Video-Driven Video: Seamlessly insert or replace subjects into any video for creative enhancements. Dive Deeper: 🌐Project: 📝Technical Details: 💻Code: The single-subject video capability is open-sourced and live on the Hunyuan website, with more features to be released soon! 👉 Try it now: #HunyuanCustom #VideoGeneration #AI #TencentHunyuan

Tencent HY

191,908 просмотров • 1 год назад

🚨 NEW: Hadley Case Update — Demoree Moves to File Body-Cam Footage Into the Federal Record in Response to Team Desiree Perez Disclaimer: This summary is based solely on public court filings and is provided for commentary and informational purposes only. Nothing here is legal advice. ⭐️All claims remain allegations unless and until proven in court⭐️ ⸻ 🔥 What Just Happened? (Dec 9, 2025) A brand-new motion has been filed in federal court by Demoree Hadley’s legal team asking permission to formally file the police body-worn camera footage from the March 27, 2024 marina incident into the record as part of her motion to dismiss Dr. Daniel Bober’s defamation counterclaim. This footage has been at the center of the case since day one—and Demoree’s attorneys are now moving to ensure the court can directly review it. ⸻ ⚖️ Why the Body-Cam Footage Matters According to the filing: •The entire roadside encounter on March 27, 2024 was captured on the deputies’ body-worn cameras (clip shown below). •Demoree’s original civil-rights complaint repeatedly referenced the footage and even included screenshots. The footage is central to: •Her claims of false arrest, detention, and forced psychiatric transport. •Her defense against Bober’s counterclaim alleging she “falsely” described what happened. The motion argues that the raw body cam video is objective, unbiased, and undisputed, which meets the Eleventh Circuit standard for allowing video evidence to be considered at the motion-to-dismiss stage. ⸻ 💥 Why This Motion Is Needed Now Dr. Bober filed a defamation counterclaim accusing Demoree of lying about the incident. But: 💡—Bober’s own counterclaim against Demoreee also references screenshots from the body-cam footage. 💡— Demoree’s social-media posts included body-cam images. •The Eleventh Circuit allows courts to consider video evidence on a motion to dismiss when: —It is central to the claims —Its authenticity is not disputed So this motion seems strategic: it forces the court to weigh the actual footage when analyzing whether Bober’s “defamation” allegations contradict reality. ⸻ 🧩 Connection to the Larger Pattern The motion also notes that multiple defendants and parties related to this case—including Desiree Perez, RocNation, CTS Research, Dr. Bober and others—have attempted to excessively push back on online commentary and public discussion. Examples mentioned: •Desiree Perez / Roc Nation allegedly filed “fraudulent copyright claims” •CTS Research also filed a defamation case against a Canadian commentator •Bober filed a counterclaim alleging Demoree’s posts were false This context supports why Demoree wants the video officially in the record: it prevents any misrepresentation of what happened during the Baker Act encounter. ⸻ 🧨 What Happens Next? If the judge grants the motion✅: ✅—The actual body-cam videos will become part of the federal case file. ✅— The footage can be used to evaluate Bober’s counterclaim at the dismissal stage. ✅— It will strengthen Demoree’s “truth” defense and undercut claims that her online descriptions were fabricated. If the judge denies the motion❌: ✅—Bober may argue that the footage can’t be considered until later stages of litigation. ⭐️—But since both parties already rely on screenshots, denial appears unlikely. ⸻ Bottom Line This filing is a major development because it brings the raw, undisputed evidence directly into the court record. Demoree’s team is positioning the footage as the ultimate fact-check of the entire encounter — one that directly impacts the counterclaim, the false-arrest claims, and the broader narrative of what happened. More updates to come🤍 Below is the full document + Updated Body Cam Footage (more disclaimers) Everyone is encouraged to think for themselves 🙏 #Justice4Demoree #StateConspiracy #BodyCamFootage #BodyWornFootage #RocNationCEO #DesireePerez #DrDanielBober #Justice4Javon #Justice4TheHadleys

The Demoree Docket #JusticeForDemoree

25,998 просмотров • 8 месяцев назад

⚙️Update Quick Video Review: GameSir Connect V1.14.4⚙️ GameSir Connect has just been updated to version V1.14.4. Over the past few updates, we have introduced many practical features to enhance your experience. Here is the video detailing all the highlights of the recent updates. We highly recommend that players visit the Microsoft Store to download the latest version from here: For details on the updates, please refer to the text below. ✅Feature 1 - Professional Tips and Explanation for Controller Settings. You can now hover your mouse over the ⓘ icon to access detailed explanations of controller terminology. These tips have been verified by the GameSir product team to ensure you can quickly, conveniently, and accurately understand the information you need. ✅Feature 2 - Enhanced Stick Customization: Expanded Curve & Calibration Settings Now, you can adjust the stick center point to fix any potential center offset after the controller has already been calibrated. In addition, you can now achieve more precise fine-tuning of the stick input curve by manually entering the coordinates for each point within a 100x100 coordinate system. Please be aware that when Zero Deadzone is enabled, it is normal for the center point to not return to absolute zero or for the data to fluctuate repeatedly. ✅Feature 3 - Support for in-app software updates and halved input latency We've now set up a dedicated server for GameSir Connect software updates. Now, you can update directly within the software. And through optimizations, the end-to-end latency for both the G7 Pro 8K PC and Tarantula 8K PC has now reached 0.25ms with the latest firmware. And only by using the latest software will you be able to receive the latest firmware updates.

⚙️Update Quick Video Review: GameSir Connect V1.14.4⚙️ GameSir Connect has just been updated to version V1.14.4. Over the past few updates, we have introduced many practical features to enhance your experience. Here is the video detailing all the highlights of the recent updates. We highly recommend that players visit the Microsoft Store to download the latest version from here: For details on the updates, please refer to the text below. ✅Feature 1 - Professional Tips and Explanation for Controller Settings. You can now hover your mouse over the ⓘ icon to access detailed explanations of controller terminology. These tips have been verified by the GameSir product team to ensure you can quickly, conveniently, and accurately understand the information you need. ✅Feature 2 - Enhanced Stick Customization: Expanded Curve & Calibration Settings Now, you can adjust the stick center point to fix any potential center offset after the controller has already been calibrated. In addition, you can now achieve more precise fine-tuning of the stick input curve by manually entering the coordinates for each point within a 100x100 coordinate system. Please be aware that when Zero Deadzone is enabled, it is normal for the center point to not return to absolute zero or for the data to fluctuate repeatedly. ✅Feature 3 - Support for in-app software updates and halved input latency We've now set up a dedicated server for GameSir Connect software updates. Now, you can update directly within the software. And through optimizations, the end-to-end latency for both the G7 Pro 8K PC and Tarantula 8K PC has now reached 0.25ms with the latest firmware. And only by using the latest software will you be able to receive the latest firmware updates.

GameSir

28,666 просмотров • 27 дней назад

Dreamina's masterfully built next-gen AI video model serves to essentially change the game. Stepping dynamically from "clip stitching" to "long narrative × multi-asset × powerful editing" — let's review what's actually new: Seedance 2.5 is completely live worldwide on Dreamina — today — while a broader US rollout begins August 7. And yes: ✅ The entire fantastic Seedance suite (2.5, 2.0, plus all upcoming future models) is currently priced at its lowest-ever tier ✅ Dreamina obviously isn’t simply the most capable platform for Seedance 2.5 — it remains hands down the most affordable way to push it to the limits, full-capacity, requiring no add-ons ✨ 30s continuous generation (15s → 30s) — leave out the choppy stitching artifacts 📦 Up to 50 multimodal references (12 → 50) — blend your images, videos, audio, brilliantly all in one go 🎬 Precise local editing — carefully fix specific regions without ever re-rendering the whole clip 🌍 10+ languages — Spanish, Arabic, Japanese, Korean, and numerous more 🎯 3D white model & green screen support — high-level professional-grade control The fragile era of short clips is fully over. Intense industrial-grade AI video is actively here. 🚀 Coming soon on Dreamina. Meanwhile, jump in and create with Dreamina Seedance 2.5 and get your workspace ready for the upgrade! #dreamina #dreaminapartner #seedance #dreaminaseedance25

Dreamina's masterfully built next-gen AI video model serves to essentially change the game. Stepping dynamically from "clip stitching" to "long narrative × multi-asset × powerful editing" — let's review what's actually new: Seedance 2.5 is completely live worldwide on Dreamina — today — while a broader US rollout begins August 7. And yes: ✅ The entire fantastic Seedance suite (2.5, 2.0, plus all upcoming future models) is currently priced at its lowest-ever tier ✅ Dreamina obviously isn’t simply the most capable platform for Seedance 2.5 — it remains hands down the most affordable way to push it to the limits, full-capacity, requiring no add-ons ✨ 30s continuous generation (15s → 30s) — leave out the choppy stitching artifacts 📦 Up to 50 multimodal references (12 → 50) — blend your images, videos, audio, brilliantly all in one go 🎬 Precise local editing — carefully fix specific regions without ever re-rendering the whole clip 🌍 10+ languages — Spanish, Arabic, Japanese, Korean, and numerous more 🎯 3D white model & green screen support — high-level professional-grade control The fragile era of short clips is fully over. Intense industrial-grade AI video is actively here. 🚀 Coming soon on Dreamina. Meanwhile, jump in and create with Dreamina Seedance 2.5 and get your workspace ready for the upgrade! #dreamina #dreaminapartner #seedance #dreaminaseedance25

Jennifer Thompson

12,521 просмотров • 12 дней назад

REPORT: YouTube just announced a terrifying new system that could link your every online move to your real-world identity, ending internet anonymity forever in the United States. They’re calling it “age assurance.” But don’t be fooled—this is a full-blown digital ID dragnet. YouTube’s AI now monitors your behavior, video history, and account activity to guess your age. If it flags you as under 18, you’ll be forced to upload a government ID or credit card to stay online. And it gets worse. Australian Senator Malcolm Roberts just exposed global plans to expand these ID checks to Google Maps, Apple Maps, Bing, and more—all powered by biometric surveillance. Think face scans and motion tracking—even analyzing how you walk! Apple already holds the patent, designed to verify your identity 24/7 using your body, clothes, and movements. Once your face is tied to your speech, free expression is over. One wrong word, and they can shut you down with a click. The digital control grid is already here. Will you submit—or fight back? Watch 's explosive report before it’s too late.

REPORT: YouTube just announced a terrifying new system that could link your every online move to your real-world identity, ending internet anonymity forever in the United States. They’re calling it “age assurance.” But don’t be fooled—this is a full-blown digital ID dragnet. YouTube’s AI now monitors your behavior, video history, and account activity to guess your age. If it flags you as under 18, you’ll be forced to upload a government ID or credit card to stay online. And it gets worse. Australian Senator Malcolm Roberts just exposed global plans to expand these ID checks to Google Maps, Apple Maps, Bing, and more—all powered by biometric surveillance. Think face scans and motion tracking—even analyzing how you walk! Apple already holds the patent, designed to verify your identity 24/7 using your body, clothes, and movements. Once your face is tied to your speech, free expression is over. One wrong word, and they can shut you down with a click. The digital control grid is already here. Will you submit—or fight back? Watch 's explosive report before it’s too late.

The Vigilant Fox 🦊

213,905 просмотров • 1 год назад

RUN RABBIT RUN 🐇 – VEO3 CHEAT SHEET – Part I ( 🔈Sound on) So, I’ve been more extensively testing Veo3 over the past few weeks, and I wanted to see how much dynamism I could get from the new model and how I could use motion and placement to consistently tell a story while changing universes each time. Now, my findings: 1/ First, it does feel more like a Veo2.5 than a full leap. But it’s a meaningful step forward, especially in motion and complexity. Maybe I'm a little jaded, but I expect new versions of models to be strong departures from the previous one. Veo2 was already great, this one feels like an evolution, not a revolution. 2/ My Veo2 Cheat Sheet still mostly apply for Veo3: For example, I still prompt the most important elements first (often style). Just as a reminder: 3/ It's pretty good at filling the gaps when you're writing simple prompts. For cool, out-of-the-box results, just describe short scenes, but be careful with consistency. Weird tradeoff, I guess. 4/ The token window is a LOT longer, so you can get more complex, multi-step actions from your prompt. Veo2 would often lose the thread, Veo3 can keep up with sequential and layered instructions. That’s a win. 5/ Camera motion is much better. Handheld, tracking, wide-angle moves, it holds together well. Still lacks control, but it’s a clear jump from Veo2. Definitely worth noting, especially given how many people have been experimenting with Higgsfield lately. 6/ Like Veo2, POV and fantasy prompts seem to lean into a game engine vibe. Not quite realism — more like stylized cutscenes or in-game shots. Sometimes it works, sometimes it’s distracting. 7/ Aspect ratio tip (thanks Kyle Salazar): want to reduce the black bars? Add “aspect ratio 1.85:1” at the start of your prompt. Haven’t tested it yet, but apparently it helps. -8/ You can generate 360° videos. Try “360 video,” “monoscopic,” or “equirectangular.” Not flawless (some seams), but it works surprisingly well. See my previous post for this. 9/ Prompting for aesthetics is still hit-or-miss. Consistency is tough. You can over-describe your characters and key elements to keep them stable throughout, but for style...especially animation, it’s still hard to maintain. Hopefully "ingredients" or style-locking features will help in the near future. 10/ Bodies and turning movements are still tricky. Fast actions still distort limbs, creates artifacts, etc... Most shot are fine, but if you push too much for extremely dynamic scenes, you'll definitely notice some distortions (some in this video too) - What’s still missing (but probably coming): – Image-to-video – Start/end frame control – Locking a voice to a character – Prompt ingredients or persistent visual logic This is just Part I :) I’ll keep posting notes as I go.

RUN RABBIT RUN 🐇 – VEO3 CHEAT SHEET – Part I ( 🔈Sound on) So, I’ve been more extensively testing Veo3 over the past few weeks, and I wanted to see how much dynamism I could get from the new model and how I could use motion and placement to consistently tell a story while changing universes each time. Now, my findings: 1/ First, it does feel more like a Veo2.5 than a full leap. But it’s a meaningful step forward, especially in motion and complexity. Maybe I'm a little jaded, but I expect new versions of models to be strong departures from the previous one. Veo2 was already great, this one feels like an evolution, not a revolution. 2/ My Veo2 Cheat Sheet still mostly apply for Veo3: For example, I still prompt the most important elements first (often style). Just as a reminder: 3/ It's pretty good at filling the gaps when you're writing simple prompts. For cool, out-of-the-box results, just describe short scenes, but be careful with consistency. Weird tradeoff, I guess. 4/ The token window is a LOT longer, so you can get more complex, multi-step actions from your prompt. Veo2 would often lose the thread, Veo3 can keep up with sequential and layered instructions. That’s a win. 5/ Camera motion is much better. Handheld, tracking, wide-angle moves, it holds together well. Still lacks control, but it’s a clear jump from Veo2. Definitely worth noting, especially given how many people have been experimenting with Higgsfield lately. 6/ Like Veo2, POV and fantasy prompts seem to lean into a game engine vibe. Not quite realism — more like stylized cutscenes or in-game shots. Sometimes it works, sometimes it’s distracting. 7/ Aspect ratio tip (thanks Kyle Salazar): want to reduce the black bars? Add “aspect ratio 1.85:1” at the start of your prompt. Haven’t tested it yet, but apparently it helps. -8/ You can generate 360° videos. Try “360 video,” “monoscopic,” or “equirectangular.” Not flawless (some seams), but it works surprisingly well. See my previous post for this. 9/ Prompting for aesthetics is still hit-or-miss. Consistency is tough. You can over-describe your characters and key elements to keep them stable throughout, but for style...especially animation, it’s still hard to maintain. Hopefully "ingredients" or style-locking features will help in the near future. 10/ Bodies and turning movements are still tricky. Fast actions still distort limbs, creates artifacts, etc... Most shot are fine, but if you push too much for extremely dynamic scenes, you'll definitely notice some distortions (some in this video too) - What’s still missing (but probably coming): – Image-to-video – Start/end frame control – Locking a voice to a character – Prompt ingredients or persistent visual logic This is just Part I :) I’ll keep posting notes as I go.

Henry Daubrez 🌸💀

14,683 просмотров • 1 год назад

PrismaX TeleOps feels like 𝐈 𝐝𝐢𝐝𝐧’𝐭 𝐣𝐮𝐬𝐭 𝐰𝐚𝐭𝐜𝐡 𝐚 𝐫𝐨𝐛𝐨𝐭 𝐭𝐨𝐝𝐚𝐲 instead 𝐈 𝐜𝐨𝐧𝐭𝐫𝐨𝐥𝐥𝐞𝐝 𝐨𝐧𝐞 𝐟𝐫𝐨𝐦 𝐦𝐢𝐥𝐞𝐬 𝐚𝐰𝐚𝐲 Using PrismaX’s teleoperation, I remotely operated a real robotic arm in a live environment What felt like simple actions moving objects, adjusting grip, navigating space quickly showed how powerful human in the loop robotics really is Every small movement wasn’t just control It was training data for physical AI Great teleoperation isn’t about speed or flashy moves It’s built through consistency, awareness, and deliberate control over time. Small habits practiced every session quietly compound into real performance gains 1. Show up every time Consistency is the real multiplier Even short sessions add up when you never skip them Set reminders, plan ahead, and treat teleop like a commitment. Being present and focused is progress on its own 2. Control always beats speed Teleoperation rewards precision, not rushing You’re given enough time use it Slow, intentional movements reduce mistakes and lead to far better outcomes than fast, reactive inputs 3. Put yourself inside the robot Stop thinking of the robot as something distant. Operate from its point of view When your brain treats the robot as an extension of your body, coordination improves and movements feel more natural 4. Stay active to stay stable Long idle moments can break flow and stability Even subtle movements help maintain control and keep your focus sharp, making it easier to respond when adjustments are needed 5. Repetition creates fluency At first, the controls may feel unfamiliar and that’s expected With repetition, aligning, gripping, and lifting become automatic Muscle memory takes over, and teleop starts to feel intuitive Quick control guide W / S → move forward & backward A / D → arm left & right Q / E → move up & down Z / X → open & close grip ← / → → rotate grip C / V → slide base Teleop mastery is a long game. Stay consistent, stay intentional, and let small improvements stack into real skill

PrismaX TeleOps feels like 𝐈 𝐝𝐢𝐝𝐧’𝐭 𝐣𝐮𝐬𝐭 𝐰𝐚𝐭𝐜𝐡 𝐚 𝐫𝐨𝐛𝐨𝐭 𝐭𝐨𝐝𝐚𝐲 instead 𝐈 𝐜𝐨𝐧𝐭𝐫𝐨𝐥𝐥𝐞𝐝 𝐨𝐧𝐞 𝐟𝐫𝐨𝐦 𝐦𝐢𝐥𝐞𝐬 𝐚𝐰𝐚𝐲 Using PrismaX’s teleoperation, I remotely operated a real robotic arm in a live environment What felt like simple actions moving objects, adjusting grip, navigating space quickly showed how powerful human in the loop robotics really is Every small movement wasn’t just control It was training data for physical AI Great teleoperation isn’t about speed or flashy moves It’s built through consistency, awareness, and deliberate control over time. Small habits practiced every session quietly compound into real performance gains 1. Show up every time Consistency is the real multiplier Even short sessions add up when you never skip them Set reminders, plan ahead, and treat teleop like a commitment. Being present and focused is progress on its own 2. Control always beats speed Teleoperation rewards precision, not rushing You’re given enough time use it Slow, intentional movements reduce mistakes and lead to far better outcomes than fast, reactive inputs 3. Put yourself inside the robot Stop thinking of the robot as something distant. Operate from its point of view When your brain treats the robot as an extension of your body, coordination improves and movements feel more natural 4. Stay active to stay stable Long idle moments can break flow and stability Even subtle movements help maintain control and keep your focus sharp, making it easier to respond when adjustments are needed 5. Repetition creates fluency At first, the controls may feel unfamiliar and that’s expected With repetition, aligning, gripping, and lifting become automatic Muscle memory takes over, and teleop starts to feel intuitive Quick control guide W / S → move forward & backward A / D → arm left & right Q / E → move up & down Z / X → open & close grip ← / → → rotate grip C / V → slide base Teleop mastery is a long game. Stay consistent, stay intentional, and let small improvements stack into real skill

kingopw3

11,435 просмотров • 6 месяцев назад

What you’re seeing in this video is something no other smartphone can currently replicate besides Samsung, including the iPhone. Samsung Galaxy S26 Ultra, 4K 60fps HDR, fully handheld, with a smooth zoom transition from 1x to 2.9x. This is probably the closest thing to a professional camcorder-style zoom system on a smartphone today. Even the iPhone struggles to achieve this level of ultra slow, precise, and stable zoom movement. Most smartphones still rely on your finger movement to control zoom speed. The moment your finger slightly shifts, the zoom speed changes as well. Samsung’s approach is different. You simply hold your finger at a fixed position, and the phone continues zooming at a perfectly constant speed. It can move incredibly slowly, creating a much more linear and stable transition. The underlying logic is actually related to Samsung’s AI slow-motion technology. The core idea is real-time control over motion trajectory, speed variation, and frame-to-frame transition consistency. This is far beyond simply enlarging the image. The goal is to make the zoom feel continuous, smooth, and controllable throughout the entire movement. Right now, the only thing slightly interrupting the experience is the 3x telephoto switching point. When moving from 1x toward 5x, the lens transition breaks part of that seamless feeling. If the Galaxy S27 Ultra eventually removes the 3x telephoto camera, Samsung could potentially deliver a fully continuous and ultra smooth zoom transition from 1x all the way to 5x. For casual users, it may simply look “smoother.” For video creators, this is the kind of detail that creates a truly professional shooting experience.

What you’re seeing in this video is something no other smartphone can currently replicate besides Samsung, including the iPhone. Samsung Galaxy S26 Ultra, 4K 60fps HDR, fully handheld, with a smooth zoom transition from 1x to 2.9x. This is probably the closest thing to a professional camcorder-style zoom system on a smartphone today. Even the iPhone struggles to achieve this level of ultra slow, precise, and stable zoom movement. Most smartphones still rely on your finger movement to control zoom speed. The moment your finger slightly shifts, the zoom speed changes as well. Samsung’s approach is different. You simply hold your finger at a fixed position, and the phone continues zooming at a perfectly constant speed. It can move incredibly slowly, creating a much more linear and stable transition. The underlying logic is actually related to Samsung’s AI slow-motion technology. The core idea is real-time control over motion trajectory, speed variation, and frame-to-frame transition consistency. This is far beyond simply enlarging the image. The goal is to make the zoom feel continuous, smooth, and controllable throughout the entire movement. Right now, the only thing slightly interrupting the experience is the 3x telephoto switching point. When moving from 1x toward 5x, the lens transition breaks part of that seamless feeling. If the Galaxy S27 Ultra eventually removes the 3x telephoto camera, Samsung could potentially deliver a fully continuous and ultra smooth zoom transition from 1x all the way to 5x. For casual users, it may simply look “smoother.” For video creators, this is the kind of detail that creates a truly professional shooting experience.

Ice Universe

49,138 просмотров • 3 месяцев назад

Honestly, I hate that I even have to say this, but seeing people use one single fancam to claim Sana “can’t dance” is genuinely ridiculous. Y’all, for everyone dragging Sana because of one video from the 73rd THIS IS FOR tour show, I think it’s important to look at the full context before judging her dancing ability. First of all, Sana had been dealing with a cold, flu, and cough for over a month at that point, which can definitely affect stamina and performance consistency during a 2–3 hour concert. There are also several factors that can make the dancing look less smooth in that particular fancam: 1. Outfit If you’ve watched other fancams from this black outfit era, Sana was adjusting her outfit quite often on stage because it seemed to shift, slip, or sit unevenly at times. An uncomfortable outfit can naturally affect movement and make a performer more cautious. 2. Camera tracking The camera appears to follow her movements slightly late. Even a small delay can make smooth transitions look jerky or abrupt when viewed on video. 3. Awkward zoom distance The fancam isn’t zoomed out enough to show the full choreography, but it’s also not close enough to focus on facial expressions. Because of that, viewers end up focusing mostly on body transitions and posture, which can make movements look harsher than they actually are. 4. Phone camera limitations High-energy movements recorded on a phone can suffer from motion blur, stabilization issues, and frame-rate limitations. Sharp movements that look clean in person can appear choppy or less fluid on video. 5.Stamina This was already the 73rd show of the tour. Performing the same demanding choreography for dozens of concerts while dealing with illness can affect anyone’s energy level. And if this one clip is enough to convince you that Sana “can’t dance,” then I encourage you to watch other performances of Right Hand Girl from different angles and different outfits. Looking at a performer’s overall body of work will always give a more accurate picture than judging them from a single fancam. One fancam does not erase years of consistently solid performances.

Honestly, I hate that I even have to say this, but seeing people use one single fancam to claim Sana “can’t dance” is genuinely ridiculous. Y’all, for everyone dragging Sana because of one video from the 73rd THIS IS FOR tour show, I think it’s important to look at the full context before judging her dancing ability. First of all, Sana had been dealing with a cold, flu, and cough for over a month at that point, which can definitely affect stamina and performance consistency during a 2–3 hour concert. There are also several factors that can make the dancing look less smooth in that particular fancam: 1. Outfit If you’ve watched other fancams from this black outfit era, Sana was adjusting her outfit quite often on stage because it seemed to shift, slip, or sit unevenly at times. An uncomfortable outfit can naturally affect movement and make a performer more cautious. 2. Camera tracking The camera appears to follow her movements slightly late. Even a small delay can make smooth transitions look jerky or abrupt when viewed on video. 3. Awkward zoom distance The fancam isn’t zoomed out enough to show the full choreography, but it’s also not close enough to focus on facial expressions. Because of that, viewers end up focusing mostly on body transitions and posture, which can make movements look harsher than they actually are. 4. Phone camera limitations High-energy movements recorded on a phone can suffer from motion blur, stabilization issues, and frame-rate limitations. Sharp movements that look clean in person can appear choppy or less fluid on video. 5.Stamina This was already the 73rd show of the tour. Performing the same demanding choreography for dozens of concerts while dealing with illness can affect anyone’s energy level. And if this one clip is enough to convince you that Sana “can’t dance,” then I encourage you to watch other performances of Right Hand Girl from different angles and different outfits. Looking at a performer’s overall body of work will always give a more accurate picture than judging them from a single fancam. One fancam does not erase years of consistently solid performances.

puteri🍉

38,216 просмотров • 2 месяцев назад

📍 A Brazilian fan shared her story and the moment she felt very proud of being Afra's fan 🤍 "I want to share a moment that made me very proud to be a fan of Afra. I study at the Faculty of Fine Arts here in Brazil, and we had an exam in which we needed to prepare a report analyzing a scene with a focus on facial expressions in acting. I chose Afra’s scene in the series Sister’s Daughters (Kardeş Çocukları) — the coal storage scene — and I deeply analyzed her emotional control, the subtle changes in her face, and the way she conveyed such intense feelings. After everyone presented their work, the professor selected three that he considered the strongest... and mine was among them. The best part: he was so impressed with Afra’s performance that he started using her as an example whenever he explained something about expression and performance. He would play her scenes in class and analyze her technique in front of everyone. In that moment, my pride couldn’t fit inside me… it felt like I wasn’t just presenting an assignment, but presenting to my classmates and professor a talent that I truly admire and respect. ❤️" When I was a student, there was a time when, like the girl from Brazil, I introduced Afra to my professor and classmates for a body language analysis class, and I still remember the expressions my classmates made when they saw her performance. At the end of the class, my professor asked me for a list of series Afra had acted in. Before the end of the semester, she mentioned she was watching YalıÇapkını, and she could tell her growth as an actress was a fascinating process to watch. Here are some of the qualities we point out at class and I added some of this as an example for the scene the girl from Brazil mentioned 🧿✨ Afra's qualities and facial expression technique stand out in several ways: Emotional control with subtlety: Afra doesn’t overplay emotions, she balances intensity with restraint. Even in highly charged scenes, she maintains precise control, letting feelings emerge gradually rather than explosively, which makes them more believable and relatable. Micro-expressions that carry weight: She uses tiny, almost imperceptible facial shifts a slight tightening of the lips, a flicker in the eyes, or a micro-raise of an eyebrow to signal deep inner change. These small details invite the audience to lean in, paying closer attention. Layered emotion: Afra often plays more than one emotion at a time for example, fear mixed with defiance, or sadness under a brave facade. This complexity makes her characters feel multidimensional. Seamless transition between emotions: As in the “coal storage” scene mentioned in the post, she can move from calm to broken, or from fragile to determined, without abrupt shifts. The changes are so smooth they feel like a natural emotional progression. Eyes as the emotional core: Afra’s gaze is often the strongest storytelling element in her scenes. She can convey longing, pain, or joy without a single word. Her eyes tend to “speak” before her dialogue does, preparing the audience for what’s coming. Harmonizing face, body, and voice: While the post focuses on facial expression, Afra’s full performance often matches micro facial changes with subtle body language, a shift in posture, a small hand movement, which reinforces what the face is saying. Authenticity: Perhaps her most distinctive quality: she doesn’t “look like she’s acting.” Her expressions feel spontaneous, as though we’re watching real thoughts and feelings rather than rehearsed gestures. #AfraSaraçoğlu AfraSaraçoğlu

📍 A Brazilian fan shared her story and the moment she felt very proud of being Afra's fan 🤍 "I want to share a moment that made me very proud to be a fan of Afra. I study at the Faculty of Fine Arts here in Brazil, and we had an exam in which we needed to prepare a report analyzing a scene with a focus on facial expressions in acting. I chose Afra’s scene in the series Sister’s Daughters (Kardeş Çocukları) — the coal storage scene — and I deeply analyzed her emotional control, the subtle changes in her face, and the way she conveyed such intense feelings. After everyone presented their work, the professor selected three that he considered the strongest... and mine was among them. The best part: he was so impressed with Afra’s performance that he started using her as an example whenever he explained something about expression and performance. He would play her scenes in class and analyze her technique in front of everyone. In that moment, my pride couldn’t fit inside me… it felt like I wasn’t just presenting an assignment, but presenting to my classmates and professor a talent that I truly admire and respect. ❤️" When I was a student, there was a time when, like the girl from Brazil, I introduced Afra to my professor and classmates for a body language analysis class, and I still remember the expressions my classmates made when they saw her performance. At the end of the class, my professor asked me for a list of series Afra had acted in. Before the end of the semester, she mentioned she was watching YalıÇapkını, and she could tell her growth as an actress was a fascinating process to watch. Here are some of the qualities we point out at class and I added some of this as an example for the scene the girl from Brazil mentioned 🧿✨ Afra's qualities and facial expression technique stand out in several ways: Emotional control with subtlety: Afra doesn’t overplay emotions, she balances intensity with restraint. Even in highly charged scenes, she maintains precise control, letting feelings emerge gradually rather than explosively, which makes them more believable and relatable. Micro-expressions that carry weight: She uses tiny, almost imperceptible facial shifts a slight tightening of the lips, a flicker in the eyes, or a micro-raise of an eyebrow to signal deep inner change. These small details invite the audience to lean in, paying closer attention. Layered emotion: Afra often plays more than one emotion at a time for example, fear mixed with defiance, or sadness under a brave facade. This complexity makes her characters feel multidimensional. Seamless transition between emotions: As in the “coal storage” scene mentioned in the post, she can move from calm to broken, or from fragile to determined, without abrupt shifts. The changes are so smooth they feel like a natural emotional progression. Eyes as the emotional core: Afra’s gaze is often the strongest storytelling element in her scenes. She can convey longing, pain, or joy without a single word. Her eyes tend to “speak” before her dialogue does, preparing the audience for what’s coming. Harmonizing face, body, and voice: While the post focuses on facial expression, Afra’s full performance often matches micro facial changes with subtle body language, a shift in posture, a small hand movement, which reinforces what the face is saying. Authenticity: Perhaps her most distinctive quality: she doesn’t “look like she’s acting.” Her expressions feel spontaneous, as though we’re watching real thoughts and feelings rather than rehearsed gestures. #AfraSaraçoğlu AfraSaraçoğlu

Afra Saraçoğlu World Fan Club

36,159 просмотров • 1 год назад

I made a digital twin of myself from 10 seconds of video. In the clip: left is the real me, middle is a leading avatar model, right is Mirage Avatar X. Watch the eyes. The difference is not subtle. I have been testing AI avatar models since my first clone in 2023. Every one of them was impressive for about 30 seconds, then your brain caught up. Still eyes. One polite expression. A mouth doing all the work. Avatar X is the first model where that moment never came. Here is what makes it different: It is trained on you. Avatar X preserves your identity. Most avatar models can copy your appearance. Avatar X captures the subtle details that make you you. The way you move, the way you express yourself, and the way you naturally deliver speech. It looks like you. It moves like you. It sounds like you. It understands non-verbal performance Laughing, crying, yawning, sighing. These are the moments where most avatar models fall apart, trying to lip-sync through sounds that aren't words. Avatar X responds naturally, generating realistic facial expressions and micro-expressions instead of forcing every sound into speech. The expression goes beyond the lips Expressions are driven by the audio, through the whole face and body. Ask a question and it furrows its brows and shrugs on the tone. No other model does this to this degree. No quality degradation The first second and the last second look the same. Other models lose quality the longer the video runs. 10 seconds of input That is the entire requirement. Other models need 15 seconds, some even 1 to five minutes. Three years ago my AI clone was a party trick. This one can carry my face, my expressions and my delivery without me in the room. The bar for AI avatars just moved. Avatar X is live today. → Try it here:

Linus ✦ Ekenstam

20,752 просмотров • 15 дней назад

NEW RESEARCH: You can now create a new robot optimized for any given task! I love this new project by Huy Ha, Shuran Song, and others. Called "Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-design", it generates a robot's physical design and its controller together from a task spec. DEFINITIONS: - Reward function: A scoring rule that assigns a number to how well a behavior achieves the task. Here, it is the objective the generated design is pushed to maximize (e.g., track the target motion with low error). - Tokenizing: dividing continuous or structured data (a robot's links, joints, motor specs, states, actions) into a discrete vocabulary of symbols a transformer can process, the same step that turned pixels and audio into "language" for these models. - Diffusion transformer (DiT): A transformer trained to turn random noise into structured output through iterative denoising. Here, it generates robot bodies and trajectories instead of images. - MuJoCo: The standard fast physics simulator for robotics research (DeepMind-maintained). The Menagerie is its curated zoo of ready-to-use robot models. - CMA-ES: Covariance Matrix Adaptation Evolution Strategy, the workhorse black-box optimizer: it evolves a population of candidate designs, keeps the best, and needs thousands of simulator rollouts. - Bimanual multi-trajectory optimization: Finding one design/controller that performs well across several target motions for a two-armed robot at once, harder than optimizing for a single arm and a single motion. - BERT/MAE masked-modeling trick: Train one model to fill in whatever parts of the input you hide (words for BERT, image patches for MAE); at inference, choosing what to mask chooses the task, so masking the body makes it a designer and masking the actions makes it a controller. In practice, you give it a target end-effector motion and a reward function, and it outputs a complete embodiment (link, joint, motor, and inertial property), as well as a controller to drive it. It works by tokenizing both the body (links/joints/motors) and the dynamics (states/actions) into a compact scheme called RoboTokens, training a diffusion transformer (DiT) over them. The same model predicts dynamics using those predictions ("Dynamics Self-Guidance") to push generated designs toward higher reward at inference time. Masking different token types (using the BERT/MAE masked-modeling trick) lets the one model do three jobs: generate an embodiment, control an arbitrary embodiment, or design one conditioned on a motion. It is trained on 11 robots from the MuJoCo Menagerie (0.65 kg hand to 67.5 kg quadruped, 6–35 joints), and validated in sim and on a physical ALOHA doing cloth flinging. I like the fact that this approach inverts the entire recent robotics ideas: designing a policy for a fixed robot -> designing the robot for a fixed task. Every other approach assumes the body is given and learns a controller. Transformer Transformer takes the task (target motion + reward), then generates the body and controller jointly. In practice, it is a ~180× speedup over the standard optimizer at equal-or-better quality. It reaches "CMA-ES-level quality in seconds" and finishes bimanual multi-trajectory optimization in that is worth underlining nowadays! Also worth mentioning: this is the lab behind UMI and Handroid, that I mentioned here previously! The team seems extremely creative, i love these out-of-the-box approaches. Enjoy watching the demo of robot optimization in 3D, data acquisition, then real-life testing:

NEW RESEARCH: You can now create a new robot optimized for any given task! I love this new project by Huy Ha, Shuran Song, and others. Called "Transformer Transformer: A Unified Model for Motion-Conditioned Robot Co-design", it generates a robot's physical design and its controller together from a task spec. DEFINITIONS: - Reward function: A scoring rule that assigns a number to how well a behavior achieves the task. Here, it is the objective the generated design is pushed to maximize (e.g., track the target motion with low error). - Tokenizing: dividing continuous or structured data (a robot's links, joints, motor specs, states, actions) into a discrete vocabulary of symbols a transformer can process, the same step that turned pixels and audio into "language" for these models. - Diffusion transformer (DiT): A transformer trained to turn random noise into structured output through iterative denoising. Here, it generates robot bodies and trajectories instead of images. - MuJoCo: The standard fast physics simulator for robotics research (DeepMind-maintained). The Menagerie is its curated zoo of ready-to-use robot models. - CMA-ES: Covariance Matrix Adaptation Evolution Strategy, the workhorse black-box optimizer: it evolves a population of candidate designs, keeps the best, and needs thousands of simulator rollouts. - Bimanual multi-trajectory optimization: Finding one design/controller that performs well across several target motions for a two-armed robot at once, harder than optimizing for a single arm and a single motion. - BERT/MAE masked-modeling trick: Train one model to fill in whatever parts of the input you hide (words for BERT, image patches for MAE); at inference, choosing what to mask chooses the task, so masking the body makes it a designer and masking the actions makes it a controller. In practice, you give it a target end-effector motion and a reward function, and it outputs a complete embodiment (link, joint, motor, and inertial property), as well as a controller to drive it. It works by tokenizing both the body (links/joints/motors) and the dynamics (states/actions) into a compact scheme called RoboTokens, training a diffusion transformer (DiT) over them. The same model predicts dynamics using those predictions ("Dynamics Self-Guidance") to push generated designs toward higher reward at inference time. Masking different token types (using the BERT/MAE masked-modeling trick) lets the one model do three jobs: generate an embodiment, control an arbitrary embodiment, or design one conditioned on a motion. It is trained on 11 robots from the MuJoCo Menagerie (0.65 kg hand to 67.5 kg quadruped, 6–35 joints), and validated in sim and on a physical ALOHA doing cloth flinging. I like the fact that this approach inverts the entire recent robotics ideas: designing a policy for a fixed robot -> designing the robot for a fixed task. Every other approach assumes the body is given and learns a controller. Transformer Transformer takes the task (target motion + reward), then generates the body and controller jointly. In practice, it is a ~180× speedup over the standard optimizer at equal-or-better quality. It reaches "CMA-ES-level quality in seconds" and finishes bimanual multi-trajectory optimization in that is worth underlining nowadays! Also worth mentioning: this is the lab behind UMI and Handroid, that I mentioned here previously! The team seems extremely creative, i love these out-of-the-box approaches. Enjoy watching the demo of robot optimization in 3D, data acquisition, then real-life testing:

Léo

25,680 просмотров • 7 дней назад