Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Controlling reaction-diffusion sim via image-texture. Can influence both initial state and parameters. #b3d #geometrynodes #reactiondiffusion

Alex Martinelli

6,153 subscribers

14,932 Aufrufe • vor 1 Jahr •via X (Twitter)

Wissenschaft & Technologie Bildung Kunst #b3d #geometrynodes #reactiondiffusion

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Reaction Diffusion in Blender Geometry-Nodes Get it for FREE on Gumroad: Throw at me your feedback and questions! #b3d #geometrynodes #reactiondiffusion

Reaction Diffusion in Blender Geometry-Nodes Get it for FREE on Gumroad: Throw at me your feedback and questions! #b3d #geometrynodes #reactiondiffusion

Alex Martinelli

13,574 Aufrufe • vor 1 Jahr

Reaction-diffusion in Blender geometry-nodes. This is the basic "single concentration blur" version. I'm also working on a stable version of the Gray-Scott model. #b3d #geometrynodes #reactiondiffusion

Reaction-diffusion in Blender geometry-nodes. This is the basic "single concentration blur" version. I'm also working on a stable version of the Gray-Scott model. #b3d #geometrynodes #reactiondiffusion

Alex Martinelli

42,888 Aufrufe • vor 1 Jahr

RIP Photoshop. Higgsfield just launched Canvas, a state-of-the-art image editing model. With just two clicks, you can swap any object in an image. Logos, text, texture and scale stay exactly as they are... Here's how it works:

RIP Photoshop. Higgsfield just launched Canvas, a state-of-the-art image editing model. With just two clicks, you can swap any object in an image. Logos, text, texture and scale stay exactly as they are... Here's how it works:

Angry Tom

156,252 Aufrufe • vor 1 Jahr

The Hidden Language of Diffusion Models paper page: tackle the challenge of understanding concept representations in text-to-image models by decomposing an input text prompt into a small set of interpretable elements. This is achieved by learning a pseudo-token that is a sparse weighted combination of tokens from the model's vocabulary, with the objective of reconstructing the images generated for the given concept. Applied over the state-of-the-art Stable Diffusion model, this decomposition reveals non-trivial and surprising structures in the representations of concepts. For example, we find that some concepts such as "a president" or "a composer" are dominated by specific instances (e.g., "Obama", "Biden") and their interpolations. Other concepts, such as "happiness" combine associated terms that can be concrete ("family", "laughter") or abstract ("friendship", "emotion"). In addition to peering into the inner workings of Stable Diffusion, our method also enables applications such as single-image decomposition to tokens, bias detection and mitigation, and semantic image manipulation

The Hidden Language of Diffusion Models paper page: tackle the challenge of understanding concept representations in text-to-image models by decomposing an input text prompt into a small set of interpretable elements. This is achieved by learning a pseudo-token that is a sparse weighted combination of tokens from the model's vocabulary, with the objective of reconstructing the images generated for the given concept. Applied over the state-of-the-art Stable Diffusion model, this decomposition reveals non-trivial and surprising structures in the representations of concepts. For example, we find that some concepts such as "a president" or "a composer" are dominated by specific instances (e.g., "Obama", "Biden") and their interpolations. Other concepts, such as "happiness" combine associated terms that can be concrete ("family", "laughter") or abstract ("friendship", "emotion"). In addition to peering into the inner workings of Stable Diffusion, our method also enables applications such as single-image decomposition to tokens, bias detection and mitigation, and semantic image manipulation

AK

41,830 Aufrufe • vor 3 Jahren

MaskINT: Video Editing via Interpolative Non-autoregressive Masked Transformers paper page: Recent advances in generative AI have significantly enhanced image and video editing, particularly in the context of text prompt control. State-of-the-art approaches predominantly rely on diffusion models to accomplish these tasks. However, the computational demands of diffusion-based methods are substantial, often necessitating large-scale paired datasets for training, and therefore challenging the deployment in practical applications. This study addresses this challenge by breaking down the text-based video editing process into two separate stages. In the first stage, we leverage an existing text-to-image diffusion model to simultaneously edit a few keyframes without additional fine-tuning. In the second stage, we introduce an efficient model called MaskINT, which is built on non-autoregressive masked generative transformers and specializes in frame interpolation between the keyframes, benefiting from structural guidance provided by intermediate frames. Our comprehensive set of experiments illustrates the efficacy and efficiency of MaskINT when compared to other diffusion-based methodologies. This research offers a practical solution for text-based video editing and showcases the potential of non-autoregressive masked generative transformers in this domain.

MaskINT: Video Editing via Interpolative Non-autoregressive Masked Transformers paper page: Recent advances in generative AI have significantly enhanced image and video editing, particularly in the context of text prompt control. State-of-the-art approaches predominantly rely on diffusion models to accomplish these tasks. However, the computational demands of diffusion-based methods are substantial, often necessitating large-scale paired datasets for training, and therefore challenging the deployment in practical applications. This study addresses this challenge by breaking down the text-based video editing process into two separate stages. In the first stage, we leverage an existing text-to-image diffusion model to simultaneously edit a few keyframes without additional fine-tuning. In the second stage, we introduce an efficient model called MaskINT, which is built on non-autoregressive masked generative transformers and specializes in frame interpolation between the keyframes, benefiting from structural guidance provided by intermediate frames. Our comprehensive set of experiments illustrates the efficacy and efficiency of MaskINT when compared to other diffusion-based methodologies. This research offers a practical solution for text-based video editing and showcases the potential of non-autoregressive masked generative transformers in this domain.

AK

25,449 Aufrufe • vor 2 Jahren

Seedance V2 This prompt can be used for any fight. I used the last two images I created with Nano banana Pro, not even a character sheet, someone could even open a new account with this prompt. The possibilities are endless. setting: location: "Ancient 'World Martial Arts Tournament' arena [@ Image 2]" details: "Clear stone platform textures, intricate Chinese guardian beast carvings, detailed ancient architecture" audio_style: "Shaw Brothers classic kung fu cinema soundtrack" action_sequence: participants: "[@ Image 1] vs [@ Image 3], both unarmed/bare-handed" choreography: opening: "[@ Image 1] moves like lightning with sharp energy-infused strikes; [@ Image 3] parries using fluid Tai Chi grandmaster techniques to neutralize the onslaught." climax: "[@ Image 1] lunges for a tail-whip ambush; [@ Image 3] counters with a powerful qi-palmed strike. [@ Image 1] dodges with ghost-like agility." finisher: "[@ Image 1] fires a Kamehameha at the chest; [@ Image 3] tanks the hit with a qi-shield and counters with a full-force palm strike, knocking [@ Image 1] off the ring." cinematography: camera: "360-degree orbital wrap-around shots, capturing every martial arts exchange" lighting: "Dynamic lighting shifts synced with combat intensity to create a tense atmosphere" visual_style: "Cinematic photorealism, 8K resolution, film-like texture" technical_quality: standard: "Low AI signature, no excessive skin smoothing, natural fluid motion" negative_constraints: "No deformed limbs, no extra/missing fingers, no clipping, no blurring, no low resolution, no cluttered backgrounds, no color banding"

Seedance V2 This prompt can be used for any fight. I used the last two images I created with Nano banana Pro, not even a character sheet, someone could even open a new account with this prompt. The possibilities are endless. setting: location: "Ancient 'World Martial Arts Tournament' arena [@ Image 2]" details: "Clear stone platform textures, intricate Chinese guardian beast carvings, detailed ancient architecture" audio_style: "Shaw Brothers classic kung fu cinema soundtrack" action_sequence: participants: "[@ Image 1] vs [@ Image 3], both unarmed/bare-handed" choreography: opening: "[@ Image 1] moves like lightning with sharp energy-infused strikes; [@ Image 3] parries using fluid Tai Chi grandmaster techniques to neutralize the onslaught." climax: "[@ Image 1] lunges for a tail-whip ambush; [@ Image 3] counters with a powerful qi-palmed strike. [@ Image 1] dodges with ghost-like agility." finisher: "[@ Image 1] fires a Kamehameha at the chest; [@ Image 3] tanks the hit with a qi-shield and counters with a full-force palm strike, knocking [@ Image 1] off the ring." cinematography: camera: "360-degree orbital wrap-around shots, capturing every martial arts exchange" lighting: "Dynamic lighting shifts synced with combat intensity to create a tense atmosphere" visual_style: "Cinematic photorealism, 8K resolution, film-like texture" technical_quality: standard: "Low AI signature, no excessive skin smoothing, natural fluid motion" negative_constraints: "No deformed limbs, no extra/missing fingers, no clipping, no blurring, no low resolution, no cluttered backgrounds, no color banding"

Emily

89,842 Aufrufe • vor 5 Monaten

NeuRBF: A Neural Fields Representation with Adaptive Radial Basis Functions paper page: present a novel type of neural fields that uses general radial bases for signal representation. State-of-the-art neural fields typically rely on grid-based representations for storing local neural features and N-dimensional linear kernels for interpolating features at continuous query points. The spatial positions of their neural features are fixed on grid nodes and cannot well adapt to target signals. Our method instead builds upon general radial bases with flexible kernel position and shape, which have higher spatial adaptivity and can more closely fit target signals. To further improve the channel-wise capacity of radial basis functions, we propose to compose them with multi-frequency sinusoid functions. This technique extends a radial basis to multiple Fourier radial bases of different frequency bands without requiring extra parameters, facilitating the representation of details. Moreover, by marrying adaptive radial bases with grid-based ones, our hybrid combination inherits both adaptivity and interpolation smoothness. We carefully designed weighting schemes to let radial bases adapt to different types of signals effectively. Our experiments on 2D image and 3D signed distance field representation demonstrate the higher accuracy and compactness of our method than prior arts. When applied to neural radiance field reconstruction, our method achieves state-of-the-art rendering quality, with small model size and comparable training speed.

NeuRBF: A Neural Fields Representation with Adaptive Radial Basis Functions paper page: present a novel type of neural fields that uses general radial bases for signal representation. State-of-the-art neural fields typically rely on grid-based representations for storing local neural features and N-dimensional linear kernels for interpolating features at continuous query points. The spatial positions of their neural features are fixed on grid nodes and cannot well adapt to target signals. Our method instead builds upon general radial bases with flexible kernel position and shape, which have higher spatial adaptivity and can more closely fit target signals. To further improve the channel-wise capacity of radial basis functions, we propose to compose them with multi-frequency sinusoid functions. This technique extends a radial basis to multiple Fourier radial bases of different frequency bands without requiring extra parameters, facilitating the representation of details. Moreover, by marrying adaptive radial bases with grid-based ones, our hybrid combination inherits both adaptivity and interpolation smoothness. We carefully designed weighting schemes to let radial bases adapt to different types of signals effectively. Our experiments on 2D image and 3D signed distance field representation demonstrate the higher accuracy and compactness of our method than prior arts. When applied to neural radiance field reconstruction, our method achieves state-of-the-art rendering quality, with small model size and comparable training speed.

AK

194,469 Aufrufe • vor 2 Jahren

EDGS: Eliminating Densification for Efficient Convergence of 3DGS Contributions: • We show that initial triangulation based on 2D correspondences can replace the incremental refinement process, fundamentally changing how 3DGS models allocate resources. • Our method reduces the path each Gaussian must travel in parameter space. Careful initialization not only accelerates convergence but also guides optimization toward a convergence point corresponding to lower reconstruction error and thus higher reconstruction quality. • Our approach outperforms both speed-optimized and quality-focused state-of-the-art models while using only half the splats of standard 3DGS. By improving initialization rather than altering the optimization process, this method is compatible with other 3DGS acceleration techniques, making it a flexible enhancement to existing models.

EDGS: Eliminating Densification for Efficient Convergence of 3DGS Contributions: • We show that initial triangulation based on 2D correspondences can replace the incremental refinement process, fundamentally changing how 3DGS models allocate resources. • Our method reduces the path each Gaussian must travel in parameter space. Careful initialization not only accelerates convergence but also guides optimization toward a convergence point corresponding to lower reconstruction error and thus higher reconstruction quality. • Our approach outperforms both speed-optimized and quality-focused state-of-the-art models while using only half the splats of standard 3DGS. By improving initialization rather than altering the optimization process, this method is compatible with other 3DGS acceleration techniques, making it a flexible enhancement to existing models.

MrNeRF

124,568 Aufrufe • vor 1 Jahr

🚨 Ajinkya Rahane revealed the story behind his decision to send Yashasvi Jaiswal off the field. 🗣️ " I sensed the situation and also I felt it was going out of hand, both the players were going out of hand and it was my responsibility to control my teammates. You respect the umpires, the officials who are on field, who are controlling the game, who are officiating the game and also the situations. Mistakes can happen to anyone, but you, as a player, you be in that limit. So, I sensed the situation, what can happen next, it was my instant reaction to send him off the field. As a player, I also felt that he must have felt bad, I said, okay, if he felt bad about me, it's okay. But I think, that point in time, it was a good thing for him. The thing which I told him about, go in from the back, put ice on your head and come back "

🚨 Ajinkya Rahane revealed the story behind his decision to send Yashasvi Jaiswal off the field. 🗣️ " I sensed the situation and also I felt it was going out of hand, both the players were going out of hand and it was my responsibility to control my teammates. You respect the umpires, the officials who are on field, who are controlling the game, who are officiating the game and also the situations. Mistakes can happen to anyone, but you, as a player, you be in that limit. So, I sensed the situation, what can happen next, it was my instant reaction to send him off the field. As a player, I also felt that he must have felt bad, I said, okay, if he felt bad about me, it's okay. But I think, that point in time, it was a good thing for him. The thing which I told him about, go in from the back, put ice on your head and come back "

OldMonkofCricket

21,540 Aufrufe • vor 28 Tagen

Everyone's sleeping on image-to-3D AI models. They can make your app look incredibly unique, with just a little effort. Here's how. This is my calorie tracker, built in a week with nothing but prompting. Just Claude Code + a couple APIs. The visuals are all AI-generated. I'll be sharing the full workflow + all the crazy technical stuff Claude and I did to make this work, so nobody has to struggle through it like me. Deep dive coming soon! Till then, this is the high-level idea: 1. Get a clean image of the food (or whatever your asset is) - In my app, the user describes foods via text, or attaches images (or both) - If text, an LLM extracts the food description and formats it into a specific prompt I tuned for this design, and we generate an image using Z-Image Turbo through fal - If image, we do the same thing but with FLUX.2 [dev] to edit the user image into our reference design - Originally, both used Google Nano Banana, but switching to open models cut costs and latency a ton 2. Gaussian splatting (2D image → 3D model) - I tried various 2D-to-3D options on fal and ended up with TripoSplat as my preferred balance of speed, cost, latency; this turns an image into a 3D model that looks super high quality (link below) - The app displays the 2D image while our backend generates the 3D splat - We "groom" the splat to reduce size and load time by culling low-opacity/scale points 3. Render efficiently on device Originally, it looked great but ran at 10 FPS. Getting to 120 FPS was a crazy journey. TL;DR: - SwiftUI had to go; it forced us to render each asset in independent MTKViews, which wasn't workable - Instead, we composite every dish into one full-bleed CAMetalLayer using MetalSplatter (link below) - We had to make some optimizations within MetalSplatter's code too, to reduce the overhead of sorting points per render Then I added some finishing touches like the subtle rotation and parallax as they move around. I think it turned out pretty cool :) Overall, this took some effort, but we still got it done in less than a day. Hopefully your agent can follow in the footsteps of mine and do it much faster. Keep an eye out for the bigger writeup, which'll give your agent everything it needs. If you have any questions, drop em below!

Everyone's sleeping on image-to-3D AI models. They can make your app look incredibly unique, with just a little effort. Here's how. This is my calorie tracker, built in a week with nothing but prompting. Just Claude Code + a couple APIs. The visuals are all AI-generated. I'll be sharing the full workflow + all the crazy technical stuff Claude and I did to make this work, so nobody has to struggle through it like me. Deep dive coming soon! Till then, this is the high-level idea: 1. Get a clean image of the food (or whatever your asset is) - In my app, the user describes foods via text, or attaches images (or both) - If text, an LLM extracts the food description and formats it into a specific prompt I tuned for this design, and we generate an image using Z-Image Turbo through fal - If image, we do the same thing but with FLUX.2 [dev] to edit the user image into our reference design - Originally, both used Google Nano Banana, but switching to open models cut costs and latency a ton 2. Gaussian splatting (2D image → 3D model) - I tried various 2D-to-3D options on fal and ended up with TripoSplat as my preferred balance of speed, cost, latency; this turns an image into a 3D model that looks super high quality (link below) - The app displays the 2D image while our backend generates the 3D splat - We "groom" the splat to reduce size and load time by culling low-opacity/scale points 3. Render efficiently on device Originally, it looked great but ran at 10 FPS. Getting to 120 FPS was a crazy journey. TL;DR: - SwiftUI had to go; it forced us to render each asset in independent MTKViews, which wasn't workable - Instead, we composite every dish into one full-bleed CAMetalLayer using MetalSplatter (link below) - We had to make some optimizations within MetalSplatter's code too, to reduce the overhead of sorting points per render Then I added some finishing touches like the subtle rotation and parallax as they move around. I think it turned out pretty cool :) Overall, this took some effort, but we still got it done in less than a day. Hopefully your agent can follow in the footsteps of mine and do it much faster. Keep an eye out for the bigger writeup, which'll give your agent everything it needs. If you have any questions, drop em below!

Anshu

19,931 Aufrufe • vor 2 Monaten

Going with the trend of viral Korean baseball video Seedance for video GPT Image 2 for image Prompt : Use the uploaded reference image as the strongest identity anchor. The woman must look like the exact same adult Japanese woman from the reference image, not a similar person. Preserve her exact facial identity, same soft oval face, same glossy lips, same delicate nose, same large expressive eyes, same pale smooth skin, and same long voluminous curly blonde hair with soft layered bangs. Create an ultra-realistic candid KBO baseball broadcast video scene during a lively night game. The woman is seated in the crowd beside her boyfriend, both wearing casual baseball jerseys among energetic cheering fans. She naturally watches the game while smiling and laughing softly with him. No dialogue, no talking, no cinematic acting. The interaction should feel completely natural and unscripted, like a real live broadcast moment accidentally captured by the stadium camera. She holds yellow cheering sticks and lightly taps them together while cheering for the team. The boyfriend leans closer and they share a very short soft kiss lasting about one second, subtle and natural, not dramatic or romanticized. Immediately after the kiss, she realizes the live stadium camera is focused on them on the giant screen. She becomes shy and slightly embarrassed for a moment, then gives a soft cute smile toward the camera while trying not to laugh. Her reaction should feel authentic, candid, and spontaneous, like a real fan unexpectedly shown on TV. Use realistic Korean baseball TV broadcast cinematography: long telephoto lens compression, slight handheld camera movement, subtle motion blur from cheering fans, realistic stadium lighting, shallow depth of field, natural skin texture, broadcast softness, authentic crowd reactions, imperfect framing, 16:9 live sports broadcast composition, genuine candid atmosphere.

Going with the trend of viral Korean baseball video Seedance for video GPT Image 2 for image Prompt : Use the uploaded reference image as the strongest identity anchor. The woman must look like the exact same adult Japanese woman from the reference image, not a similar person. Preserve her exact facial identity, same soft oval face, same glossy lips, same delicate nose, same large expressive eyes, same pale smooth skin, and same long voluminous curly blonde hair with soft layered bangs. Create an ultra-realistic candid KBO baseball broadcast video scene during a lively night game. The woman is seated in the crowd beside her boyfriend, both wearing casual baseball jerseys among energetic cheering fans. She naturally watches the game while smiling and laughing softly with him. No dialogue, no talking, no cinematic acting. The interaction should feel completely natural and unscripted, like a real live broadcast moment accidentally captured by the stadium camera. She holds yellow cheering sticks and lightly taps them together while cheering for the team. The boyfriend leans closer and they share a very short soft kiss lasting about one second, subtle and natural, not dramatic or romanticized. Immediately after the kiss, she realizes the live stadium camera is focused on them on the giant screen. She becomes shy and slightly embarrassed for a moment, then gives a soft cute smile toward the camera while trying not to laugh. Her reaction should feel authentic, candid, and spontaneous, like a real fan unexpectedly shown on TV. Use realistic Korean baseball TV broadcast cinematography: long telephoto lens compression, slight handheld camera movement, subtle motion blur from cheering fans, realistic stadium lighting, shallow depth of field, natural skin texture, broadcast softness, authentic crowd reactions, imperfect framing, 16:9 live sports broadcast composition, genuine candid atmosphere.

Synthia

11,860 Aufrufe • vor 3 Monaten

The kitchen's open 🍕 What slice are you? Take the quiz → get your flavour → post your card → flex. We're giving away 5 BNB Chain swag boxes to randomly selected winners to celebrate Bitcoin Pizza Day. Here's how to win: 1️⃣ Take the “What Slice Are You?” quiz and get your pizza card here: 2️⃣ Quote tweet this post with your card image, #BNBPizzaDay, and tag BNB Chain. Terms: 🔸Competition period: May 22 – May 28. 🔸Prize distribution: 5 BNB Chain swag boxes, one per winner. 🔸Winners: 5 winners picked via Simpliers and announced May 28. Prizes shipped after winners confirm shipping details. 🔸Note: must complete both steps above to be eligible to win. 🔸Full terms and info can be found on our blog:

The kitchen's open 🍕 What slice are you? Take the quiz → get your flavour → post your card → flex. We're giving away 5 BNB Chain swag boxes to randomly selected winners to celebrate Bitcoin Pizza Day. Here's how to win: 1️⃣ Take the “What Slice Are You?” quiz and get your pizza card here: 2️⃣ Quote tweet this post with your card image, #BNBPizzaDay, and tag BNB Chain. Terms: 🔸Competition period: May 22 – May 28. 🔸Prize distribution: 5 BNB Chain swag boxes, one per winner. 🔸Winners: 5 winners picked via Simpliers and announced May 28. Prizes shipped after winners confirm shipping details. 🔸Note: must complete both steps above to be eligible to win. 🔸Full terms and info can be found on our blog:

BNB Chain

216,693 Aufrufe • vor 3 Monaten

🚨 SCIENTISTS JUST DISCOVERED THAT BACTERIA HAVE BEEN USING A QUANTUM ENERGY TRICK FOR BILLIONS OF YEARS. In a breakthrough published in Nature Chemistry, researchers from the University of Sheffield (working with IBM and international collaborators) have shown that purple photosynthetic bacteria (Rhodobacter sphaeroides) use singlet fission (SF) as a functional mechanism to enhance light harvesting. When carotenoids absorb blue-green light (via their S₂ state), the energy rapidly undergoes heterofission: the singlet exciton splits into a pair of triplet excitons shared between a carotenoid and its neighboring bacteriochlorophyll (BChl) a molecule. These triplet pairs are long-lived and weakly coupled. They then recombine via triplet-triplet annihilation (TTA), efficiently populating the BChl a Qy singlet state which then transfers energy to the reaction center. This SF-mediated pathway contributes up to ~18% of total carotenoid-to-BChl energy transfer in some complexes and introduces a temporal “buffering” effect, staggering energy arrival at the reaction center. Why this matters: • It is the clearest demonstration yet that singlet fission plays a genuine, functional role in natural photosynthesis • The heterofission mechanism (triplets on adjacent Crt and BChl molecules) explains how bacteria achieve ultrafast SF while maintaining long-lived, harvestable triplet pairs • This pathway helps organisms make better use of blue-green light and may protect the reaction center during fluctuating light conditions The deeper implication: Nature has evolved a sophisticated way to temporarily store and then cooperatively release electronic energy using spin-protected triplet states. This biological strategy offers new inspiration for designing artificial light-harvesting and energy-conversion systems that can buffer and regulate energy flow something synthetic SF materials have struggled to achieve. We are learning that evolution solved some of the hardest problems in photophysics long before we started trying. How might understanding biological singlet fission change the way we design future solar energy or quantum-inspired technologies? Follow for more frontier photosynthesis research and natural quantum biology.

🚨 SCIENTISTS JUST DISCOVERED THAT BACTERIA HAVE BEEN USING A QUANTUM ENERGY TRICK FOR BILLIONS OF YEARS. In a breakthrough published in Nature Chemistry, researchers from the University of Sheffield (working with IBM and international collaborators) have shown that purple photosynthetic bacteria (Rhodobacter sphaeroides) use singlet fission (SF) as a functional mechanism to enhance light harvesting. When carotenoids absorb blue-green light (via their S₂ state), the energy rapidly undergoes heterofission: the singlet exciton splits into a pair of triplet excitons shared between a carotenoid and its neighboring bacteriochlorophyll (BChl) a molecule. These triplet pairs are long-lived and weakly coupled. They then recombine via triplet-triplet annihilation (TTA), efficiently populating the BChl a Qy singlet state which then transfers energy to the reaction center. This SF-mediated pathway contributes up to ~18% of total carotenoid-to-BChl energy transfer in some complexes and introduces a temporal “buffering” effect, staggering energy arrival at the reaction center. Why this matters: • It is the clearest demonstration yet that singlet fission plays a genuine, functional role in natural photosynthesis • The heterofission mechanism (triplets on adjacent Crt and BChl molecules) explains how bacteria achieve ultrafast SF while maintaining long-lived, harvestable triplet pairs • This pathway helps organisms make better use of blue-green light and may protect the reaction center during fluctuating light conditions The deeper implication: Nature has evolved a sophisticated way to temporarily store and then cooperatively release electronic energy using spin-protected triplet states. This biological strategy offers new inspiration for designing artificial light-harvesting and energy-conversion systems that can buffer and regulate energy flow something synthetic SF materials have struggled to achieve. We are learning that evolution solved some of the hardest problems in photophysics long before we started trying. How might understanding biological singlet fission change the way we design future solar energy or quantum-inspired technologies? Follow for more frontier photosynthesis research and natural quantum biology.

TheNewPhysics

75,999 Aufrufe • vor 2 Monaten

We used our agent to push Seedance 2.5’s photorealistic skin rendering to the limit. Right from the opening shot, you can clearly see the natural texture of the skin. As the video plays, notice the subtle reflections on the glasses and the natural movement of each strand of hair. We’ve also improved both the canvas and the agent, making the entire creative process simpler than ever. Full prompt below: @ Image1 is the sole visual reference for the adult woman’s identity, facial structure, hairstyle, thin round metal glasses, black ribbed cardigan, silver cross necklace, indoor background, lighting, and selfie composition. Create a 15-second vertical 9:16 TikTok-style selfie video filmed with a front-facing smartphone camera. Preserve her exact identity, facial proportions, hairstyle, glasses, wardrobe, necklace, skin tone, and relaxed refined appearance throughout the entire video. She holds the phone at the same arm’s-length distance and slightly tilted angle shown in the reference image. Keep the mood natural, intimate, confident, and tasteful. 0–3 seconds | close-up | slightly tilted high-angle selfie perspective matching Image | 26mm equivalent front-camera lens | handheld selfie movement: she looks directly into the lens with a calm expression, subtly adjusts her phone grip, and forms a soft half-smile. Keep her eyes, lips, glasses, and facial surface in sharp natural focus. Soft window light should reveal visible but subtle natural pores, fine vellus hair, delicate under-eye texture, realistic lip lines, slight tonal variation, and natural facial asymmetry. 3–7 seconds | medium close-up | same arm’s-length selfie angle, gently rotated a few degrees to her left | 28mm equivalent | slow handheld arc: she tilts her head slightly toward the window and lets a few loose strands of hair fall near her cheek. Preserve realistic glasses reflections, detailed hair flyaways, natural skin texture, and subtle breathing. The movement should feel like an authentic smartphone selfie rather than a staged fashion shoot. 7–11 seconds | close-up | slightly lower flattering selfie angle while maintaining the same phone-held perspective | 30mm equivalent | gentle handheld push-in as she brings the phone slightly closer. She briefly glances to the side, then returns her gaze to the lens with a relaxed, quietly alluring expression. Maintain consistent facial identity and stable fine detail across frames, including natural pores, fine facial hairs, soft under-eye texture, realistic eyelid creases, natural lip texture, and small micro-expressions. 11–15 seconds | close-up | return to the original reference angle | 26mm equivalent | gentle handheld settle: she lightly touches a strand of hair near her cheek, holds eye contact, and ends with a restrained warm smile. Keep her skin naturally detailed and softly illuminated, with realistic pore visibility and fine facial texture that remains consistent during movement. End naturally without freezing into an artificial pose. Use soft directional daylight from the left side of the frame, muted neutral colors, realistic smartphone front-camera exposure, subtle handheld micro-movement, natural focus breathing, physically accurate reflections on the glasses, and authentic facial motion. The skin should look unfiltered and lifelike, with visible but subtle pores and fine detail, never airbrushed or artificially sharpened. Include quiet indoor room tone, faint fabric movement, and soft phone-handling sounds. No dialogue, no music, no captions, no text overlays. Negative constraints: do not change her identity, facial structure, age, hairstyle, glasses, clothing, necklace, background, or camera perspective. No beauty-filter smoothing, poreless skin, waxy complexion, excessive sharpening, artificial HDR, harsh skin texture, exaggerated wrinkles, gritty complexion, acne emphasis, unstable pores, crawling facial details, skin flicker, changing skin tone, face warping, eye asymmetry, melted glasses, duplicated jewelry, extra fingers, distorted hands, unnatural head rotations, sudden zooms, third-person camera angles, hard cuts, flickering exposure, overexposure, underexposure, artificial poses, or explicit content.

We used our agent to push Seedance 2.5’s photorealistic skin rendering to the limit. Right from the opening shot, you can clearly see the natural texture of the skin. As the video plays, notice the subtle reflections on the glasses and the natural movement of each strand of hair. We’ve also improved both the canvas and the agent, making the entire creative process simpler than ever. Full prompt below: @ Image1 is the sole visual reference for the adult woman’s identity, facial structure, hairstyle, thin round metal glasses, black ribbed cardigan, silver cross necklace, indoor background, lighting, and selfie composition. Create a 15-second vertical 9:16 TikTok-style selfie video filmed with a front-facing smartphone camera. Preserve her exact identity, facial proportions, hairstyle, glasses, wardrobe, necklace, skin tone, and relaxed refined appearance throughout the entire video. She holds the phone at the same arm’s-length distance and slightly tilted angle shown in the reference image. Keep the mood natural, intimate, confident, and tasteful. 0–3 seconds | close-up | slightly tilted high-angle selfie perspective matching Image | 26mm equivalent front-camera lens | handheld selfie movement: she looks directly into the lens with a calm expression, subtly adjusts her phone grip, and forms a soft half-smile. Keep her eyes, lips, glasses, and facial surface in sharp natural focus. Soft window light should reveal visible but subtle natural pores, fine vellus hair, delicate under-eye texture, realistic lip lines, slight tonal variation, and natural facial asymmetry. 3–7 seconds | medium close-up | same arm’s-length selfie angle, gently rotated a few degrees to her left | 28mm equivalent | slow handheld arc: she tilts her head slightly toward the window and lets a few loose strands of hair fall near her cheek. Preserve realistic glasses reflections, detailed hair flyaways, natural skin texture, and subtle breathing. The movement should feel like an authentic smartphone selfie rather than a staged fashion shoot. 7–11 seconds | close-up | slightly lower flattering selfie angle while maintaining the same phone-held perspective | 30mm equivalent | gentle handheld push-in as she brings the phone slightly closer. She briefly glances to the side, then returns her gaze to the lens with a relaxed, quietly alluring expression. Maintain consistent facial identity and stable fine detail across frames, including natural pores, fine facial hairs, soft under-eye texture, realistic eyelid creases, natural lip texture, and small micro-expressions. 11–15 seconds | close-up | return to the original reference angle | 26mm equivalent | gentle handheld settle: she lightly touches a strand of hair near her cheek, holds eye contact, and ends with a restrained warm smile. Keep her skin naturally detailed and softly illuminated, with realistic pore visibility and fine facial texture that remains consistent during movement. End naturally without freezing into an artificial pose. Use soft directional daylight from the left side of the frame, muted neutral colors, realistic smartphone front-camera exposure, subtle handheld micro-movement, natural focus breathing, physically accurate reflections on the glasses, and authentic facial motion. The skin should look unfiltered and lifelike, with visible but subtle pores and fine detail, never airbrushed or artificially sharpened. Include quiet indoor room tone, faint fabric movement, and soft phone-handling sounds. No dialogue, no music, no captions, no text overlays. Negative constraints: do not change her identity, facial structure, age, hairstyle, glasses, clothing, necklace, background, or camera perspective. No beauty-filter smoothing, poreless skin, waxy complexion, excessive sharpening, artificial HDR, harsh skin texture, exaggerated wrinkles, gritty complexion, acne emphasis, unstable pores, crawling facial details, skin flicker, changing skin tone, face warping, eye asymmetry, melted glasses, duplicated jewelry, extra fingers, distorted hands, unnatural head rotations, sudden zooms, third-person camera angles, hard cuts, flickering exposure, overexposure, underexposure, artificial poses, or explicit content.

underwood

30,574 Aufrufe • vor 16 Tagen

🧬 We have many foundation models or language models for DNAs, but can we control them? We introduce Ctrl-DNA: Controllable Cell-Type-Specific Regulatory DNA Design via Constrained RL — a reinforcement learning framework for controllable cis-regulatory sequence generation. Paper: Code: 🔬What’s the challenge? Designing regulatory DNA that is both highly expressive in target cell types and inactive in others is essential for synthetic biology, gene therapy, and precision medicine. Yet, controlling these trade-offs is challenging due to sparse, sequence-level rewards and biological constraints. 🔥Why Ctrl-DNA? Ctrl-DNA fine-tunes pre-trained DNA language models using a value model free, Lagrangian-guided RL framework, enabling flexible and customizable constraint optimization. Users can define application-specific thresholds across cell types, balancing expression strength with specificity. ✅ Maximize target-cell expression ✅ Constrain off-target activity under user-defined thresholds ✅ Preserve cell-type-specific TF motif structure Benchmarked on human enhancer and promoter datasets, Ctrl-DNA consistently outperforms prior methods, achieving stronger specificity, higher fitness, and more biologically grounded sequence generation — all with direct control over regulatory trade-offs. Shoutout to the PhD students Xingyu Chen (Xingyu Chen ) and Rex Ma (Rex Ma) for their amazing work leading this project!

🧬 We have many foundation models or language models for DNAs, but can we control them? We introduce Ctrl-DNA: Controllable Cell-Type-Specific Regulatory DNA Design via Constrained RL — a reinforcement learning framework for controllable cis-regulatory sequence generation. Paper: Code: 🔬What’s the challenge? Designing regulatory DNA that is both highly expressive in target cell types and inactive in others is essential for synthetic biology, gene therapy, and precision medicine. Yet, controlling these trade-offs is challenging due to sparse, sequence-level rewards and biological constraints. 🔥Why Ctrl-DNA? Ctrl-DNA fine-tunes pre-trained DNA language models using a value model free, Lagrangian-guided RL framework, enabling flexible and customizable constraint optimization. Users can define application-specific thresholds across cell types, balancing expression strength with specificity. ✅ Maximize target-cell expression ✅ Constrain off-target activity under user-defined thresholds ✅ Preserve cell-type-specific TF motif structure Benchmarked on human enhancer and promoter datasets, Ctrl-DNA consistently outperforms prior methods, achieving stronger specificity, higher fitness, and more biologically grounded sequence generation — all with direct control over regulatory trade-offs. Shoutout to the PhD students Xingyu Chen (Xingyu Chen ) and Rex Ma (Rex Ma) for their amazing work leading this project!

Bo Wang

30,719 Aufrufe • vor 1 Jahr

Birdsong is not just sound. It is data made physical. If you could see the air at the exact moment a hemp bunting sings, you would not see empty space. You would see a structured three-dimensional data array. What we hear as a soft “chirp” can be mapped as frequencies, rhythms, amplitudes, and harmonic relationships. A scatter plot turns a fleeting song into a topographic map of sound. And the technical beauty is remarkable: → Sound is a mechanical wave, built from compressions and rarefactions in the air. → The bird controls it through the syrinx, a vocal organ capable of generating two frequencies at once. → Frequency shapes pitch. → Amplitude shapes volume and cluster density. → Timbre creates the unique waveform, the texture of each “sound island.” → Each cluster shows the acoustic proximity of syllables and motifs. What looks like chaos is not chaos. It is a bioengineered signal. Territory. Genetic profile. Hormonal state. Aggression level. Mating fitness. All encoded into patterns of pitch, timing, volume, and timbre. This is why I find it so fascinating. A bird is not simply singing into the air. It is organizing the air. It is carving space into sectors of influence using sound pressure. It is making a physical claim to territory. For some, birdsong is peaceful background music. For others, it is a complex mathematical model calibrated by millions of years of evolution for survival. And here is the urgent lesson for the AI age: We often mistake invisible systems for simplicity. A bird sings, and we hear romance. An AI responds, and we see magic. But underneath both are signals, compression, feedback loops, optimization, and information architecture. The future belongs to those who can read what others dismiss as noise. So I’ll ask you: When you hear birdsong, do you hear music, data, or both? #AI #ArtificialIntelligence #Bioacoustics #Nature #Technology #Data #MachineLearning #Innovation #FutureOfWork #Signals #Evolution

Pascal Bornet

11,376 Aufrufe • vor 4 Monaten

SITREP: University of North Carolina Wilmington reportedly on Lockdown, Possible Multiple Gunmen • The University of North Carolina at Wilmington (UNCW) initiated a campus-wide lockdown at approximately 7:20 PM EDT, following unconfirmed reports of a person with a firearm on campus. • University Police (UPD) issued an immediate shelter-in-place alert via the campus notification system, advising students, faculty, and staff to seek indoor shelter, lock doors and windows, and avoid external areas. • No confirmed shots fired have been reported, and there are no known injuries. Multiple law enforcement agencies, including local Wilmington PD and state partners, are on scene conducting sweeps. • Initial reports described a single individual spotted with a gun near a parking lot or campus building (exact location unspecified in alerts). • Social media videos circulating show a figure walking with what appears to be a firearm, though authenticity is unverified. • Claims on social media mentions "multiple gunmen," "shots fired," "militia activity," or "Proud Boys threats," but as far as I can tell now, there’s only a single unconfirmed gunman sighting.

SITREP: University of North Carolina Wilmington reportedly on Lockdown, Possible Multiple Gunmen • The University of North Carolina at Wilmington (UNCW) initiated a campus-wide lockdown at approximately 7:20 PM EDT, following unconfirmed reports of a person with a firearm on campus. • University Police (UPD) issued an immediate shelter-in-place alert via the campus notification system, advising students, faculty, and staff to seek indoor shelter, lock doors and windows, and avoid external areas. • No confirmed shots fired have been reported, and there are no known injuries. Multiple law enforcement agencies, including local Wilmington PD and state partners, are on scene conducting sweeps. • Initial reports described a single individual spotted with a gun near a parking lot or campus building (exact location unspecified in alerts). • Social media videos circulating show a figure walking with what appears to be a firearm, though authenticity is unverified. • Claims on social media mentions "multiple gunmen," "shots fired," "militia activity," or "Proud Boys threats," but as far as I can tell now, there’s only a single unconfirmed gunman sighting.

MJTruthUltra

31,025 Aufrufe • vor 11 Monaten

Bitlight Wallet & RGB Faucet - Support Feature Announcement We have added a "Support" button to the RGB Faucet page to help efficiently address issues users may encounter during usage. There are two main categories of issues: 1. Claim Failed Users may sometimes not receive assets after initiating a claim. This can happen when a user performs both a claim and transfer operation simultaneously, causing a UTXO state change that results in the claimed asset being burned. In such cases, the Faucet may display the claim as Confirmed, but the user does not actually receive the asset. In rare situations, the asset may also fail to be successfully sent by the Agent. 2. RGB Assets Become Invisible After Transfer This issue requires the user to log in with the Sender's wallet, and provide the Receiver’s xpub address along with the amount transferred between the two parties. There are two common causes: a) Consignment Issue: The Sender’s wallet shows normal status and a corresponding "sent" record, but the Receiver does not receive the asset. If the Receiver has not changed their UTXO status, the asset can typically be restored and made visible. If the UTXO status has changed, the asset may be burned and cannot be recovered. b) Pay Timeout Issue: When the Sender attempts to send the asset, the wallet may get stuck in an infinite loading state, resulting in an abnormal asset state on the Sender’s side, and the Receiver also fails to receive the asset. How We Will Help For both categories of issues, we will analyze the situation based on the information you submit via the Support form. We will do our best to resolve your issue within the limitations of the RGB protocol. Due to the single-use-seal nature of the RGB protocol, if your issue is caused by a UTXO state change that results in the asset being burned, we will not be able to recover the asset. Please ensure that all information submitted is accurate and truthful. Providing incorrect information may result in your issue being unresolvable.

Bitlight Wallet & RGB Faucet - Support Feature Announcement We have added a "Support" button to the RGB Faucet page to help efficiently address issues users may encounter during usage. There are two main categories of issues: 1. Claim Failed Users may sometimes not receive assets after initiating a claim. This can happen when a user performs both a claim and transfer operation simultaneously, causing a UTXO state change that results in the claimed asset being burned. In such cases, the Faucet may display the claim as Confirmed, but the user does not actually receive the asset. In rare situations, the asset may also fail to be successfully sent by the Agent. 2. RGB Assets Become Invisible After Transfer This issue requires the user to log in with the Sender's wallet, and provide the Receiver’s xpub address along with the amount transferred between the two parties. There are two common causes: a) Consignment Issue: The Sender’s wallet shows normal status and a corresponding "sent" record, but the Receiver does not receive the asset. If the Receiver has not changed their UTXO status, the asset can typically be restored and made visible. If the UTXO status has changed, the asset may be burned and cannot be recovered. b) Pay Timeout Issue: When the Sender attempts to send the asset, the wallet may get stuck in an infinite loading state, resulting in an abnormal asset state on the Sender’s side, and the Receiver also fails to receive the asset. How We Will Help For both categories of issues, we will analyze the situation based on the information you submit via the Support form. We will do our best to resolve your issue within the limitations of the RGB protocol. Due to the single-use-seal nature of the RGB protocol, if your issue is caused by a UTXO state change that results in the asset being burned, we will not be able to recover the asset. Please ensure that all information submitted is accurate and truthful. Providing incorrect information may result in your issue being unresolvable.

Bitlight Wallet⚡️

17,044 Aufrufe • vor 1 Jahr

🚨 SCIENTISTS JUST REPLACED EXPENSIVE PLATINUM CATALYSTS IN ZINC-AIR BATTERIES WITH CHEAP IRON AND IT WORKS BETTER. Zinc-air batteries are one of the most promising low-cost, high-energy-density alternatives to lithium-ion, but they’ve been held back by a slow and inefficient oxygen reduction reaction that usually requires precious metal catalysts. Researchers at Tohoku University engineered a simple iron oxide/samarium oxide interface that dramatically speeds up this reaction. The heterointerface changes how electrons behave at the surface, weakens excessive bonding with reaction intermediates, and delivers faster kinetics plus excellent durability all without any noble metals. The new catalyst performed strongly in both liquid and flexible solid-state zinc-air batteries, successfully powering LEDs and even charging a smartphone. Why this matters: • Zinc-air batteries use oxygen from the air, abundant zinc, and are much cheaper and safer than lithium-ion • Removing the need for platinum or other precious metals makes them far more scalable and affordable • The iron-based interface approach is simple, stable in alkaline conditions, and improves both performance and longevity • It was demonstrated in practical devices, not just lab tests The deeper implication: We’re getting closer to energy storage that doesn’t rely on scarce, expensive materials. Zinc-air technology has long been held back by catalyst limitations, but this work shows that clever interface engineering with cheap, abundant elements can unlock the performance needed for real-world applications from portable electronics to large-scale grid storage. It’s another step toward clean energy systems that are not only sustainable in operation, but also in the materials they use. The future of batteries may not depend on mining more rare metals… but on smarter chemistry with what we already have in abundance. How close do you think we are to zinc-air batteries becoming a mainstream alternative to lithium-ion? Follow for more frontier battery materials and clean energy storage research.

🚨 SCIENTISTS JUST REPLACED EXPENSIVE PLATINUM CATALYSTS IN ZINC-AIR BATTERIES WITH CHEAP IRON AND IT WORKS BETTER. Zinc-air batteries are one of the most promising low-cost, high-energy-density alternatives to lithium-ion, but they’ve been held back by a slow and inefficient oxygen reduction reaction that usually requires precious metal catalysts. Researchers at Tohoku University engineered a simple iron oxide/samarium oxide interface that dramatically speeds up this reaction. The heterointerface changes how electrons behave at the surface, weakens excessive bonding with reaction intermediates, and delivers faster kinetics plus excellent durability all without any noble metals. The new catalyst performed strongly in both liquid and flexible solid-state zinc-air batteries, successfully powering LEDs and even charging a smartphone. Why this matters: • Zinc-air batteries use oxygen from the air, abundant zinc, and are much cheaper and safer than lithium-ion • Removing the need for platinum or other precious metals makes them far more scalable and affordable • The iron-based interface approach is simple, stable in alkaline conditions, and improves both performance and longevity • It was demonstrated in practical devices, not just lab tests The deeper implication: We’re getting closer to energy storage that doesn’t rely on scarce, expensive materials. Zinc-air technology has long been held back by catalyst limitations, but this work shows that clever interface engineering with cheap, abundant elements can unlock the performance needed for real-world applications from portable electronics to large-scale grid storage. It’s another step toward clean energy systems that are not only sustainable in operation, but also in the materials they use. The future of batteries may not depend on mining more rare metals… but on smarter chemistry with what we already have in abundance. How close do you think we are to zinc-air batteries becoming a mainstream alternative to lithium-ion? Follow for more frontier battery materials and clean energy storage research.

TheNewPhysics

36,798 Aufrufe • vor 2 Monaten