Загрузка видео...

Не удалось загрузить видео

На главную

Latent mask inpainting with H3 enables highly divergent changes while holding consistency. Examples by PhoenixRisen + AbleJones - workflow below.

banodoco's profile picture

Banodoco

4,269 subscribers

10,288 просмотров • 23 дней назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

76 five-second clips exploring different animation styles with MiniMax H3. All generated locally on a six-year-old GPU by the_shadow_nyc:
6:32

76 five-second clips exploring different animation styles with MiniMax H3. All generated locally on a six-year-old GPU by the_shadow_nyc:

Banodoco

31,735 просмотров • 1 месяц назад

New work with Yide Shentu! We present Latent Code as Bridges🌉 or LCB. We leverage LLMs for robotics by finetuning the LLM to output latent embeddings to control the lower level policy. This hierarchical approach enables end2end training while maintaining the advantages of LLMs!
0:24

New work with Yide Shentu! We present Latent Code as Bridges🌉 or LCB. We leverage LLMs for robotics by finetuning the LLM to output latent embeddings to control the lower level policy. This hierarchical approach enables end2end training while maintaining the advantages of LLMs!

Philipp Wu

21,085 просмотров • 2 лет назад

It's been a crazy couple of days, trying all sorts of things with ComfyUI + AnimateDiff-Evolved and various ControlNets. I'm absolutely blown away by the consistency and clarity of the animations using this method. Workflow below 👇
1:08

It's been a crazy couple of days, trying all sorts of things with ComfyUI + AnimateDiff-Evolved and various ControlNets. I'm absolutely blown away by the consistency and clarity of the animations using this method. Workflow below 👇

Purz.ai

49,496 просмотров • 2 лет назад

World models have a causality problem. Realistic videos are not enough. A world model should predict the future caused by an action, not just a plausible future. We find that many latent-action world models generate convincing videos while barely responding to the supplied action. The root cause lies in Latent Action Models (LAMs): reconstruction objectives entangle action-induced dynamics with action-irrelevant visual changes, producing visually confounded latent actions. Our solution is CD-LAM (Causally Debiased Latent Action Model). It cleanses latent action representations before pre-training begins, requiring zero changes to backbone architectures, latent dimensions, or action interfaces. Key Benchmark Results: - Action Controllability: Cuts action-following error by >30% while simultaneously enhancing visual fidelity. - Extreme Sample Efficiency: Achieves baseline performance in just 3,000 post-training steps instead of 50,000 - a >10x speedup. By removing visual confounding, CD-LAM bridges the gap between passive video prediction and actionable causal intelligence for robotics. The next bottleneck for world models isn’t realism. It’s causality. Paper, project page and additional resources in the reply below. #WorldModels #EmbodiedAI #Robotics #CausalAI
0:53

World models have a causality problem. Realistic videos are not enough. A world model should predict the future caused by an action, not just a plausible future. We find that many latent-action world models generate convincing videos while barely responding to the supplied action. The root cause lies in Latent Action Models (LAMs): reconstruction objectives entangle action-induced dynamics with action-irrelevant visual changes, producing visually confounded latent actions. Our solution is CD-LAM (Causally Debiased Latent Action Model). It cleanses latent action representations before pre-training begins, requiring zero changes to backbone architectures, latent dimensions, or action interfaces. Key Benchmark Results: - Action Controllability: Cuts action-following error by >30% while simultaneously enhancing visual fidelity. - Extreme Sample Efficiency: Achieves baseline performance in just 3,000 post-training steps instead of 50,000 - a >10x speedup. By removing visual confounding, CD-LAM bridges the gap between passive video prediction and actionable causal intelligence for robotics. The next bottleneck for world models isn’t realism. It’s causality. Paper, project page and additional resources in the reply below. #WorldModels #EmbodiedAI #Robotics #CausalAI

Biwei Huang

14,907 просмотров • 1 месяц назад

Outpaint IC-LoRA for LTX Video 2.3-22b by Mohamed Oumoumad -- extend any video's canvas by marking regions in black. Sides, top, bottom, any combo. The model fills it in with temporally consistent content.
0:43

Outpaint IC-LoRA for LTX Video 2.3-22b by Mohamed Oumoumad -- extend any video's canvas by marking regions in black. Sides, top, bottom, any combo. The model fills it in with temporally consistent content.

Banodoco

21,315 просмотров • 5 месяцев назад

i've been making a sonic.exe asym for a while, so time i show gameplay of SONIC.EXE voiced by Dansheno🎙️ we are hiring musicians! reply below if you wanna apply with examples of your work full video showing off everything in replies (pls dont flop) #sonicexe #execommunity
2:22

i've been making a sonic.exe asym for a while, so time i show gameplay of SONIC.EXE voiced by Dansheno🎙️ we are hiring musicians! reply below if you wanna apply with examples of your work full video showing off everything in replies (pls dont flop) #sonicexe #execommunity

Oc_TheGrunt

25,292 просмотров • 1 месяц назад

Battle of the Bulge: Winter Hell Around Bastogne (1944) 🇺🇸🇩🇪 In December 1944, U.S. soldiers fought around Bastogne during Germany's Ardennes Offensive. With temperatures far below zero, they endured frostbite, exhaustion, and days in snow-filled foxholes while holding the line against the German offensive. One of the toughest and most brutal campaigns fought by American forces on the Western Front.
1:15

Battle of the Bulge: Winter Hell Around Bastogne (1944) 🇺🇸🇩🇪 In December 1944, U.S. soldiers fought around Bastogne during Germany's Ardennes Offensive. With temperatures far below zero, they endured frostbite, exhaustion, and days in snow-filled foxholes while holding the line against the German offensive. One of the toughest and most brutal campaigns fought by American forces on the Western Front.

War Times

26,725 просмотров • 2 месяцев назад

ScenarioControl 🚗🛣️ - Scenario Generation from a single Dashcam Image 📸 or Text Prompt 💬!! Excited to introduce a new vision-language control mechanism for learned driving scenario generation. Given a single dashcam image or a scene prompt or an image, we generate a full scene layout 🧩, temporally consistent rollouts, including map 🗺️, agents 🚗, and ego video🛣️ ScenarioControl enables direct, fine-grained control over layout and traffic while preserving realism. It operates in a vectorized latent space with a new cross-global control mechanism to fuse vision-language inputs with scene structure while preserving realism. Interfaces seamlessly with generative video models! Project: Super fun project by Lili Gao, Yanbo Xu , William Koch, Samuele Ruffino, Luke Rowe , Behdad Chalaki, Dmitriy Rivkin, Julian Ost, Roger Girgis, Mario Bijelic.
0:52

ScenarioControl 🚗🛣️ - Scenario Generation from a single Dashcam Image 📸 or Text Prompt 💬!! Excited to introduce a new vision-language control mechanism for learned driving scenario generation. Given a single dashcam image or a scene prompt or an image, we generate a full scene layout 🧩, temporally consistent rollouts, including map 🗺️, agents 🚗, and ego video🛣️ ScenarioControl enables direct, fine-grained control over layout and traffic while preserving realism. It operates in a vectorized latent space with a new cross-global control mechanism to fuse vision-language inputs with scene structure while preserving realism. Interfaces seamlessly with generative video models! Project: Super fun project by Lili Gao, Yanbo Xu , William Koch, Samuele Ruffino, Luke Rowe , Behdad Chalaki, Dmitriy Rivkin, Julian Ost, Roger Girgis, Mario Bijelic.

Felix Heide

22,361 просмотров • 4 месяцев назад

🚨SCOTUS Issues Major EMERGENCY MIDTERM RULING The Supreme Court just delivered a massive 8-1 ruling that slashes political ad rates by nearly 50% ahead of the 2026 midterms, dealing a crippling blow to the Democrats. With the GOP holding a staggering $1 billion war chest while the DNC sits in debt, this legal victory changes the entire political battlefield. Here is the breakdown of the decision, the numbers, and why this is a total game changer. Follow: Morse Report
10:20

🚨SCOTUS Issues Major EMERGENCY MIDTERM RULING The Supreme Court just delivered a massive 8-1 ruling that slashes political ad rates by nearly 50% ahead of the 2026 midterms, dealing a crippling blow to the Democrats. With the GOP holding a staggering $1 billion war chest while the DNC sits in debt, this legal victory changes the entire political battlefield. Here is the breakdown of the decision, the numbers, and why this is a total game changer. Follow: Morse Report

Morse Report

37,041 просмотров • 3 дней назад

self portait gaussian splat I captured this by myself with a single camera. I mounted the cam, sat on an office chair, and swiveled around a full 360 while recording a video (check next tweet). Used RealityScan's AI masking to mask myself from the surroundings (for sfm to work), then used MrNeRF's LichtFeldStudio to splat! alignment is faaar from perfect (notice the ghosting/doubling), but not a bad result for such a simple workflow! and of course, Codex and I wrote a end-to-end video to splat script to repeate this as many times as I'd like :) view on supersplat:
0:40

self portait gaussian splat I captured this by myself with a single camera. I mounted the cam, sat on an office chair, and swiveled around a full 360 while recording a video (check next tweet). Used RealityScan's AI masking to mask myself from the surroundings (for sfm to work), then used MrNeRF's LichtFeldStudio to splat! alignment is faaar from perfect (notice the ghosting/doubling), but not a bad result for such a simple workflow! and of course, Codex and I wrote a end-to-end video to splat script to repeate this as many times as I'd like :) view on supersplat:

Azad Balabanian

32,732 просмотров • 6 месяцев назад

Today, we're introducing AlphaEarth Foundations from Google DeepMind , an AI model that functions like a virtual satellite which helps scientists make informed decisions on critical issues like food security, deforestation, and water resources. AlphaEarth Foundations provides a powerful new lens for understanding our planet by solving two major challenges: data overload and inconsistent information. 1️⃣ It combines information from dozens of sources to analyze the world's land and coastal waters in 10x10 meter squares, allowing for remarkable precision while tracking changes over time. 2️⃣ The system's key innovation is creating a highly compact summary for each square. These summaries require 16x less storage than those produced by other AI systems and enables scientists to create detailed, consistent maps of our planet, on-demand. AlphaEarth Foundations represents a significant step forward in understanding the state and dynamics of our changing planet. 🌎🌎🌎
0:38

Today, we're introducing AlphaEarth Foundations from Google DeepMind , an AI model that functions like a virtual satellite which helps scientists make informed decisions on critical issues like food security, deforestation, and water resources. AlphaEarth Foundations provides a powerful new lens for understanding our planet by solving two major challenges: data overload and inconsistent information. 1️⃣ It combines information from dozens of sources to analyze the world's land and coastal waters in 10x10 meter squares, allowing for remarkable precision while tracking changes over time. 2️⃣ The system's key innovation is creating a highly compact summary for each square. These summaries require 16x less storage than those produced by other AI systems and enables scientists to create detailed, consistent maps of our planet, on-demand. AlphaEarth Foundations represents a significant step forward in understanding the state and dynamics of our changing planet. 🌎🌎🌎

Google AI

130,307 просмотров • 1 год назад

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."
7:17

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

MTS

55,187 просмотров • 1 месяц назад

Creating an experimental anime film with Seedance 2.5 on CapCut has completely changed how I approach AI video production. For this project, I started with a character sheet to lock the character’s identity, costume, facial features, color palette, and hand-drawn ink style. From there, I built cinematic sequences around extreme close-ups, dynamic camera movement, expressive ink strokes, motion smears, typography, and editorial-style compositions. What makes this workflow especially powerful is that Seedance 2.5 is officially available on CapCut, giving creators first access to its full capabilities directly inside the creative workflow. Instead of generating a shot and moving to another tool for everything else, I can generate with Seedance 2.5 and continue refining the result in the same workflow and selecting the strongest moments, refining individual shots, making localized edits, and directing changes with natural-language commands. The long-form capability is also important for projects like this. Being able to create shots up to 30 seconds and extend scenes while maintaining character, scene, and motion consistency makes it possible to think beyond isolated AI clips and start building actual visual sequences. For this project, precision matters. Timestamp-based prompting, reference-based generation, localized editing, and more controlled creative direction allow me to design the sequence almost like a traditional production pipeline, while keeping the speed and flexibility of generative AI. The result I’m aiming for isn’t simply “an AI-generated anime video.” It’s a cohesive visual language: hand-drawn emerald ink, editorial Japanese typography, expressive animation, cinematic framing, and a consistent character moving through an evolving graphic world. This is where AI video becomes much more interesting to me, not just generating images that move, but directing an entire visual production from concept to final sequence. Seedance 2.5 + CapCut gives creators a much more complete environment to explore that workflow. And this is only the beginning. Powered by : CapCut CapCut Indonesia
1:00

Creating an experimental anime film with Seedance 2.5 on CapCut has completely changed how I approach AI video production. For this project, I started with a character sheet to lock the character’s identity, costume, facial features, color palette, and hand-drawn ink style. From there, I built cinematic sequences around extreme close-ups, dynamic camera movement, expressive ink strokes, motion smears, typography, and editorial-style compositions. What makes this workflow especially powerful is that Seedance 2.5 is officially available on CapCut, giving creators first access to its full capabilities directly inside the creative workflow. Instead of generating a shot and moving to another tool for everything else, I can generate with Seedance 2.5 and continue refining the result in the same workflow and selecting the strongest moments, refining individual shots, making localized edits, and directing changes with natural-language commands. The long-form capability is also important for projects like this. Being able to create shots up to 30 seconds and extend scenes while maintaining character, scene, and motion consistency makes it possible to think beyond isolated AI clips and start building actual visual sequences. For this project, precision matters. Timestamp-based prompting, reference-based generation, localized editing, and more controlled creative direction allow me to design the sequence almost like a traditional production pipeline, while keeping the speed and flexibility of generative AI. The result I’m aiming for isn’t simply “an AI-generated anime video.” It’s a cohesive visual language: hand-drawn emerald ink, editorial Japanese typography, expressive animation, cinematic framing, and a consistent character moving through an evolving graphic world. This is where AI video becomes much more interesting to me, not just generating images that move, but directing an entire visual production from concept to final sequence. Seedance 2.5 + CapCut gives creators a much more complete environment to explore that workflow. And this is only the beginning. Powered by : CapCut CapCut Indonesia

Kick Nost

11,258 просмотров • 19 дней назад

A common Pakistani confronted the so-called BYC sit-in in Islamabad with a simple question: “Why are you hiding your faces?” That question tore the mask off the entire protest. The reality is that these 20-25 people aren’t peaceful demonstrators, they are known members of the BLA and BYC, funded by RAW and Mossad, staging a fake rights movement to push a separatist agenda. Women holding placards for “missing persons” are being used as political cover while the masked men behind them are part of armed groups responsible for terror in Balochistan. This is not activism. This is narrative warfare. And the Govt. Of #Pakistan & #Balochistan must respond accordingly by exposing, dismantling and neutralising this foreign-funded theatre. #BLAisKillingBalochs #MahrangAgentOfChaos
2:17

A common Pakistani confronted the so-called BYC sit-in in Islamabad with a simple question: “Why are you hiding your faces?” That question tore the mask off the entire protest. The reality is that these 20-25 people aren’t peaceful demonstrators, they are known members of the BLA and BYC, funded by RAW and Mossad, staging a fake rights movement to push a separatist agenda. Women holding placards for “missing persons” are being used as political cover while the masked men behind them are part of armed groups responsible for terror in Balochistan. This is not activism. This is narrative warfare. And the Govt. Of #Pakistan & #Balochistan must respond accordingly by exposing, dismantling and neutralising this foreign-funded theatre. #BLAisKillingBalochs #MahrangAgentOfChaos

Hamza Lashari

18,654 просмотров • 1 год назад

Trump just laid out a brutal truth about Iran that mainstream media is too afraid to talk about, but it’s what happens after his warning that changes everything. While the world was distracted by ceasefire headlines, a secret geopolitical chess move by the US and Oman completely stripped Iran of its ultimate leverage in the Strait of Hormuz—effectively trapping them in a corner with nowhere left to turn. They thought they were holding all the cards, but this hidden maneuver forces them into a terrifying ultimatum: either surrender their control forever, or do the one thing that will inevitably trigger a devastating, all-out global war... Watch the full episode on my YT channel! Link is in the comments 📷
2:47

Trump just laid out a brutal truth about Iran that mainstream media is too afraid to talk about, but it’s what happens after his warning that changes everything. While the world was distracted by ceasefire headlines, a secret geopolitical chess move by the US and Oman completely stripped Iran of its ultimate leverage in the Strait of Hormuz—effectively trapping them in a corner with nowhere left to turn. They thought they were holding all the cards, but this hidden maneuver forces them into a terrifying ultimatum: either surrender their control forever, or do the one thing that will inevitably trigger a devastating, all-out global war... Watch the full episode on my YT channel! Link is in the comments 📷

Jillian Michaels

93,527 просмотров • 2 месяцев назад

As previously indicated, we will now draw our attention toward Soldier and Weapon Quality of Life Improvements arriving as part of #Battlefield2042 Season 6: Dark Creations. One thing we wanted to address with Update 6.0 was something our Community had highlighted to us via feedback. We had become aware of a crosshair and recoil inconsistency with high fire rate weaponry where the crosshair would take a moment to 'catch up' to bullet recoil trajectory under sustained fire. Update 6.0 makes improvements towards this through a series of changes to recoil, timings, etc. Which now improves the feeling and consistency of all weapons when compensating for recoil and bursting. Additionally, we have changed almost all weaponry, even the ones least affected by this, to tighten the "pull" sensation which makes the camera track recoil trajectory. This will result in the crosshair staying much closer to the center at all times while we maintain the feel of recoil movement expressed through the camera. Thank you for your continued feedback! Stay tuned for further Soldier & Weapon Quality of Life Improvements, and full 6.0 Update Notes soon! 👇Here are a few early before/after examples. Content Warning; Lots of brrrrts.
0:59

As previously indicated, we will now draw our attention toward Soldier and Weapon Quality of Life Improvements arriving as part of #Battlefield2042 Season 6: Dark Creations. One thing we wanted to address with Update 6.0 was something our Community had highlighted to us via feedback. We had become aware of a crosshair and recoil inconsistency with high fire rate weaponry where the crosshair would take a moment to 'catch up' to bullet recoil trajectory under sustained fire. Update 6.0 makes improvements towards this through a series of changes to recoil, timings, etc. Which now improves the feeling and consistency of all weapons when compensating for recoil and bursting. Additionally, we have changed almost all weaponry, even the ones least affected by this, to tighten the "pull" sensation which makes the camera track recoil trajectory. This will result in the crosshair staying much closer to the center at all times while we maintain the feel of recoil movement expressed through the camera. Thank you for your continued feedback! Stay tuned for further Soldier & Weapon Quality of Life Improvements, and full 6.0 Update Notes soon! 👇Here are a few early before/after examples. Content Warning; Lots of brrrrts.

Battlefield Comms

741,881 просмотров • 2 лет назад

Today, we introduce HunyuanImage 3.0-Instruct, a native multimodal model focusing on image-editing by integrating visual understanding with precise image synthesis! 🚀 It understands input images and reasons before generating images. Built on an 80B-parameter MoE architecture (13B activated), it natively unifies deep multimodal comprehension and high-fidelity generation. 🧠 A "Thinking" Model with Native CoT & MixGRPO: The model doesn’t just execute commands, it processes them through a Native Chain-of-Thought (CoT) schema. Enhanced by our self-developed MixGRPO algorithm, it reasons through complex instructions to achieve flawless intent alignment and human-preference consistency. 🎨 Precise Editing & Multi-Image Fusion: The model enables accurate image editing by adding, removing, or modifying elements while keeping non-target areas perfectly intact. It also excels at seamless multi-image fusion, synthesizing complex scenes by extracting and blending elements from multiple sources into a unified, consistent output. 🏆 SOTA Performance: HunyuanImage 3.0-Instruct sets a new benchmark in visual quality and alignment, delivering performance that matches leading proprietary models. We aim to enable the community to explore new ideas with a state-of-the-art foundation model, fostering a dynamic and vibrant image generation ecosystem. 🛠️🎨 💻Try it at (PC only):
1:12

Today, we introduce HunyuanImage 3.0-Instruct, a native multimodal model focusing on image-editing by integrating visual understanding with precise image synthesis! 🚀 It understands input images and reasons before generating images. Built on an 80B-parameter MoE architecture (13B activated), it natively unifies deep multimodal comprehension and high-fidelity generation. 🧠 A "Thinking" Model with Native CoT & MixGRPO: The model doesn’t just execute commands, it processes them through a Native Chain-of-Thought (CoT) schema. Enhanced by our self-developed MixGRPO algorithm, it reasons through complex instructions to achieve flawless intent alignment and human-preference consistency. 🎨 Precise Editing & Multi-Image Fusion: The model enables accurate image editing by adding, removing, or modifying elements while keeping non-target areas perfectly intact. It also excels at seamless multi-image fusion, synthesizing complex scenes by extracting and blending elements from multiple sources into a unified, consistent output. 🏆 SOTA Performance: HunyuanImage 3.0-Instruct sets a new benchmark in visual quality and alignment, delivering performance that matches leading proprietary models. We aim to enable the community to explore new ideas with a state-of-the-art foundation model, fostering a dynamic and vibrant image generation ecosystem. 🛠️🎨 💻Try it at (PC only):

Tencent Hy

126,668 просмотров • 7 месяцев назад