Загрузка видео...

Не удалось загрузить видео

На главную

DIFFUSOR dropping this week! ⚔️ Convolution based Diffusion using a special ir generator with custom windowing, minimal comb filtering, random seeds and latency compensation! 🥳

Naskomusic's profile picture

Nasko

11,936 subscribers

17,860 просмотров • 3 месяцев 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

guys guys!!! i can't!! Fable 5 is going crazy. I made this flowing ink painting canvas on Happycapy using Fable 5. Soft ink blooms, slow drifting strokes, and pigment dissolving like color in water. Built with Three.js + WebGL fluid simulation, pointer-based ink injection, soft diffusion, low turbulence, and a minimal interactive canvas. a 1:1 recreation, prompt in thread ↓
0:28

guys guys!!! i can't!! Fable 5 is going crazy. I made this flowing ink painting canvas on Happycapy using Fable 5. Soft ink blooms, slow drifting strokes, and pigment dissolving like color in water. Built with Three.js + WebGL fluid simulation, pointer-based ink injection, soft diffusion, low turbulence, and a minimal interactive canvas. a 1:1 recreation, prompt in thread ↓

Min Zhou

19,732 просмотров • 2 месяцев назад

Diffusion Policy was one of the big breakthroughs that has enabled an explosion in real-world robot learning. However, it’s always had a weakness, which is that it works by computing a final action trajectory from random noise, which leads to high latency when predicting a final action sequence. Instead, why not initialize the search based on previous actions? This allows for incredibly fast policy inference and in many cases improved generalization, generating high-quality predictions with sub-ms latency. Jindou Jia and Jianfei Yang join us to explain. Learn more on Episode 95 of RoboPapers, with Michael Cho - Rbt/Acc and Chris Paxton!
49:19

Diffusion Policy was one of the big breakthroughs that has enabled an explosion in real-world robot learning. However, it’s always had a weakness, which is that it works by computing a final action trajectory from random noise, which leads to high latency when predicting a final action sequence. Instead, why not initialize the search based on previous actions? This allows for incredibly fast policy inference and in many cases improved generalization, generating high-quality predictions with sub-ms latency. Jindou Jia and Jianfei Yang join us to explain. Learn more on Episode 95 of RoboPapers, with Michael Cho - Rbt/Acc and Chris Paxton!

RoboPapers

28,155 просмотров • 28 дней назад

I think I went a bit too hard with this documentation thing... Fully custom canvas based UI with scrolling, text wrapping and optimized draw calls, using Aseprite's API David Capello would be proud of me ehe #Asemulator
0:25

I think I went a bit too hard with this documentation thing... Fully custom canvas based UI with scrolling, text wrapping and optimized draw calls, using Aseprite's API David Capello would be proud of me ehe #Asemulator

Goomboo

15,757 просмотров • 7 месяцев назад

GIVEAWAY ANNOUNCEMENT In collab with Gotta_Custom_Em_All, leader in custom cards, we are dropping a Giveaway tomorrow, Monday 10am Eastern, May 25th, Memorial Day, for a truly special, custom Peak Hobby slab: Pikachu with a Straw Hat! Follow both of us Gotta_Custom_Em_All and Peak Hobby, Like & Repost to get alerted when this Giveaway drops!
0:20

GIVEAWAY ANNOUNCEMENT In collab with Gotta_Custom_Em_All, leader in custom cards, we are dropping a Giveaway tomorrow, Monday 10am Eastern, May 25th, Memorial Day, for a truly special, custom Peak Hobby slab: Pikachu with a Straw Hat! Follow both of us Gotta_Custom_Em_All and Peak Hobby, Like & Repost to get alerted when this Giveaway drops!

Peak Hobby

59,974 просмотров • 3 месяцев назад

Today is my birthday, and I’m excited to celebrate with you all! As a special treat, here’s a sneak peek of my new music video dropping this week. Enjoy! 🎶🔥 #NewMusic #BirthdayCelebration #SneakPeek #StayTuned
0:38

Today is my birthday, and I’m excited to celebrate with you all! As a special treat, here’s a sneak peek of my new music video dropping this week. Enjoy! 🎶🔥 #NewMusic #BirthdayCelebration #SneakPeek #StayTuned

Yolanda Adams

36,595 просмотров • 2 лет назад

First AI-Powered Blink Generator is Live! 👁️🤖 Chat with our AI Agent and generate your custom Blink in under a minute — now live for your Donations! 👉 The first 500 users who generate Blinks with our AI Agent and share on X will receive a special prize. 🐈 Details in this 🧵
0:36

First AI-Powered Blink Generator is Live! 👁️🤖 Chat with our AI Agent and generate your custom Blink in under a minute — now live for your Donations! 👉 The first 500 users who generate Blinks with our AI Agent and share on X will receive a special prize. 🐈 Details in this 🧵

WAVE 🌊

60,749 просмотров • 1 год назад

releasing sakura v1.0 a high-performance minimal (~1k lines of code) terminal-based multimedia library that renders images, gifs, and videos using sixel graphics with synchronized audio playback. sakura supports all modern formats: jpg, png, bmp, animated gifs, mp4, avi, mov and mkv with adaptive scaling and advanced frame timing and skipping for smooth playback. it also supports custom image processing and batch processing with decent-ish log output. more examples in replies:
1:43

releasing sakura v1.0 a high-performance minimal (~1k lines of code) terminal-based multimedia library that renders images, gifs, and videos using sixel graphics with synchronized audio playback. sakura supports all modern formats: jpg, png, bmp, animated gifs, mp4, avi, mov and mkv with adaptive scaling and advanced frame timing and skipping for smooth playback. it also supports custom image processing and batch processing with decent-ish log output. more examples in replies:

sλrthak

181,004 просмотров • 1 год назад

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.
5:11

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

AK

83,657 просмотров • 3 лет назад

Positional tracking on smart glasses and webcams Experimenting with Niantic Spatial 🌎’s 8th Wall engine to get position using camera and motion data Works with compatible smartglasses, in this case a modified Omi AI’s Omi Glass (based on the Seeed Studio ESP32-S3 Sense) running custom firmware Also works on webcams using our motion modules for supplying imu data
1:13

Positional tracking on smart glasses and webcams Experimenting with Niantic Spatial 🌎’s 8th Wall engine to get position using camera and motion data Works with compatible smartglasses, in this case a modified Omi AI’s Omi Glass (based on the Seeed Studio ESP32-S3 Sense) running custom firmware Also works on webcams using our motion modules for supplying imu data

Brilliant Wear

14,800 просмотров • 5 месяцев назад

Realtime-VLA FLASH tackles one of the biggest deployment bottlenecks for diffusion-based VLAs: inference latency. The key idea is speculative inference for flow-matching VLAs. A lightweight draft model predicts an action chunk, while the main model’s Action Expert verifies it in parallel using flow-consistency checks instead of running full denoising every replanning round. This lets the system replace many expensive 58 ms full inference rounds with speculative rounds as fast as 7.8 ms, reducing average latency to 19.1 ms and achieving a 3.04× speedup on LIBERO while largely preserving success rate. Interesting systems insight: they profile π0 and show VLM prefill is compute-bound, while Action Denoise is memory-bound. FLASH exploits this by reusing KV cache and parallelizing verification instead of repeatedly running sequential denoising.
1:54

Realtime-VLA FLASH tackles one of the biggest deployment bottlenecks for diffusion-based VLAs: inference latency. The key idea is speculative inference for flow-matching VLAs. A lightweight draft model predicts an action chunk, while the main model’s Action Expert verifies it in parallel using flow-consistency checks instead of running full denoising every replanning round. This lets the system replace many expensive 58 ms full inference rounds with speculative rounds as fast as 7.8 ms, reducing average latency to 19.1 ms and achieving a 3.04× speedup on LIBERO while largely preserving success rate. Interesting systems insight: they profile π0 and show VLM prefill is compute-bound, while Action Denoise is memory-bound. FLASH exploits this by reusing KV cache and parallelizing verification instead of repeatedly running sequential denoising.

Robots Digest 🤖

14,576 просмотров • 3 месяцев назад

LegoGPT, an LLM-based system that generates physically stable LEGO structures from text prompts, backed by a new 47,000+ sample dataset and physics-aware filtering during inference. → LegoGPT is trained on a custom dataset, StableText2Lego, which includes 47,000+ 3D LEGO models mapped to text, spanning 28,000+ unique objects. → The model predicts LEGO bricks sequentially like tokens, using next-token prediction in a transformer setup. → To ensure physical stability, LegoGPT integrates physics-aware rollback and validity filtering, pruning out structurally invalid brick placements. → The generated designs are aesthetically aligned with prompts, physically buildable, and tested both with human manual assembly and robotic arms. → The team also introduced a text-driven LEGO coloring/texturing pipeline, enabling more expressive and customized outputs. → The dataset, code, and models are all publicly released under an open-access license.
2:10

LegoGPT, an LLM-based system that generates physically stable LEGO structures from text prompts, backed by a new 47,000+ sample dataset and physics-aware filtering during inference. → LegoGPT is trained on a custom dataset, StableText2Lego, which includes 47,000+ 3D LEGO models mapped to text, spanning 28,000+ unique objects. → The model predicts LEGO bricks sequentially like tokens, using next-token prediction in a transformer setup. → To ensure physical stability, LegoGPT integrates physics-aware rollback and validity filtering, pruning out structurally invalid brick placements. → The generated designs are aesthetically aligned with prompts, physically buildable, and tested both with human manual assembly and robotic arms. → The team also introduced a text-driven LEGO coloring/texturing pipeline, enabling more expressive and customized outputs. → The dataset, code, and models are all publicly released under an open-access license.

Rohan Paul

75,281 просмотров • 1 год назад

Three 16 year old Indian students can remove microplastics from water using tamarind seeds. And they've received a $12,500 grant from The Earth Prize to develop this solution. How did they do it? Tamarind seeds have something called polysaccharide - which is a sticky compound that can natural polymer. Kinda like a plant-based glue. This glue when mixed with water can pull out microplastics and clump them together. And these kids are using the same principle to develop their product called "Plas-Stick" - which is just this power made from waste tamarind seeds. Now, "Plas-Stick" is still in its developmental stage and there is a lot that needs to be done before it can be scaled. But if they pull this off - we could have a simple, biodegradable and scalable solution to purify water for billions of people in the world. And that's something. Kudos to Vivaan Chhawchharia, Ariana Agarwal, and Avyana Mehta. P.S. Video and Picture taken from The Earth Prize
1:16

Three 16 year old Indian students can remove microplastics from water using tamarind seeds. And they've received a $12,500 grant from The Earth Prize to develop this solution. How did they do it? Tamarind seeds have something called polysaccharide - which is a sticky compound that can natural polymer. Kinda like a plant-based glue. This glue when mixed with water can pull out microplastics and clump them together. And these kids are using the same principle to develop their product called "Plas-Stick" - which is just this power made from waste tamarind seeds. Now, "Plas-Stick" is still in its developmental stage and there is a lot that needs to be done before it can be scaled. But if they pull this off - we could have a simple, biodegradable and scalable solution to purify water for billions of people in the world. And that's something. Kudos to Vivaan Chhawchharia, Ariana Agarwal, and Avyana Mehta. P.S. Video and Picture taken from The Earth Prize

Varun Guru

62,876 просмотров • 3 месяцев назад

STEAL THE BRAINROT is expanding. We’re hiring a high-level UEFN Level Designer to lead our in-game events. We’re looking for someone very experienced and creative, who knows how to bring their talent and ideas into high-quality in-game events. There is an existing structure and visual style, and the role requires the ability to adapt and build new events based on what has already been done. Requirements: - Several years of experience in level / environment design - Strong portfolio - All assets must be custom made - Strong experience with Blender - Deep understanding of materials - Advanced UEFN experience (materials, VFX, optimization) - Full-time availability - Available to start immediately We’re looking for talented people with real experience, not beginners. Compensation is high and based on experience. A signing bonus is available for the right candidate. Below are examples of past events, this is the quality bar: 100% custom assets, modeled in Blender, with custom materials & VFX in UEFN, fully optimized. DM me with your portfolio.
0:41

STEAL THE BRAINROT is expanding. We’re hiring a high-level UEFN Level Designer to lead our in-game events. We’re looking for someone very experienced and creative, who knows how to bring their talent and ideas into high-quality in-game events. There is an existing structure and visual style, and the role requires the ability to adapt and build new events based on what has already been done. Requirements: - Several years of experience in level / environment design - Strong portfolio - All assets must be custom made - Strong experience with Blender - Deep understanding of materials - Advanced UEFN experience (materials, VFX, optimization) - Full-time availability - Available to start immediately We’re looking for talented people with real experience, not beginners. Compensation is high and based on experience. A signing bonus is available for the right candidate. Below are examples of past events, this is the quality bar: 100% custom assets, modeled in Blender, with custom materials & VFX in UEFN, fully optimized. DM me with your portfolio.

FeRinS

137,307 просмотров • 8 месяцев назад

Our OSS engineer Clelia Bertelli (🦀/acc) recently built 𝗹𝗶𝘁𝗲𝘀𝗲𝗮𝗿𝗰𝗵, a fully local document ingestion and retrieval CLI/TUI application powered by LiteParse ⚡ litesearch demonstrates how developers can assemble a high-performance, local-first retrieval pipeline using open tools from across the ecosystem: • Parsing: LiteParse, the fast and accurate document parser we recently open sourced • Chunking: @ChonkieAI • Embeddings: A local Nomic model via Hugging Face transformers.js • Vector storage: A local Qdrant edge shard (custom-built in Rust and compiled as a native add-on) • Retrieval: Query stored files with optional path-based filtering and configurable relevance thresholds • Runtime: Bun for speed and versatility 💻 Check out the repository and try it yourself: 📚 LiteParse docs:
0:32

Our OSS engineer Clelia Bertelli (🦀/acc) recently built 𝗹𝗶𝘁𝗲𝘀𝗲𝗮𝗿𝗰𝗵, a fully local document ingestion and retrieval CLI/TUI application powered by LiteParse ⚡ litesearch demonstrates how developers can assemble a high-performance, local-first retrieval pipeline using open tools from across the ecosystem: • Parsing: LiteParse, the fast and accurate document parser we recently open sourced • Chunking: @ChonkieAI • Embeddings: A local Nomic model via Hugging Face transformers.js • Vector storage: A local Qdrant edge shard (custom-built in Rust and compiled as a native add-on) • Retrieval: Query stored files with optional path-based filtering and configurable relevance thresholds • Runtime: Bun for speed and versatility 💻 Check out the repository and try it yourself: 📚 LiteParse docs:

LlamaIndex 🦙

14,900 просмотров • 5 месяцев назад

In the past few weeks, I deep dived into an exploration revolving around the use of physical interfaces to feed and interact with a real-time img2img diffusion pipeline using Stream Diffusion and SDXL Turbo. What really captivated me is to use my hands, objects, art supplies, tools, and light to create images and scenes. 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗠𝗲𝗱𝗶𝗮 + 𝗧𝗼𝗼𝗹𝘀 I experimented with clay, manipulating different types and colors along with a selection of prompts. I used a magnifying glass, tracked in real-time, to focus the diffusion process on specific areas. Combining these tools created a dynamic and inspiring experience. Using magic clay to layer shapes and colors as a base for revealing landscapes and hidden worlds, and the magnifying glass to focus and reveal these details, was particularly effective. 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗟𝗶𝗴𝗵𝘁 I used light as my method of interaction with the img2img diffusion. This approach felt special right away. There was something magical about holding a physical light source and seeing it influence the generated visuals. I iterated on this technique with themes like Rococo architecture, flowers, Brutalist architecture, hidden worlds, and origami landscapes. 𝗜𝗻𝗸 + 𝗛𝘆𝗯𝗿𝗶𝗱 𝗙𝗼𝗿𝗺𝗮𝘁𝘀 I also used ink in milk as a means of physical interaction with the diffusion pipeline. As I drop ink into milk, shapes come alive instantly. By learning to manipulate the combination of physical and digital elements, I steered the generated output toward my areas of interest. These iterations extended beyond ink in milk to include the format in which these elements are contained: a circular plate or a triptych of small stainless steel trays. These formats provide a structured yet flexible framework to explore themes and narratives across multiple visual spaces. It's magical. Some of that last iteration has been captured in this insightful article by Fast Company: #stablediffusion #realtime #ai
1:11

In the past few weeks, I deep dived into an exploration revolving around the use of physical interfaces to feed and interact with a real-time img2img diffusion pipeline using Stream Diffusion and SDXL Turbo. What really captivated me is to use my hands, objects, art supplies, tools, and light to create images and scenes. 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗠𝗲𝗱𝗶𝗮 + 𝗧𝗼𝗼𝗹𝘀 I experimented with clay, manipulating different types and colors along with a selection of prompts. I used a magnifying glass, tracked in real-time, to focus the diffusion process on specific areas. Combining these tools created a dynamic and inspiring experience. Using magic clay to layer shapes and colors as a base for revealing landscapes and hidden worlds, and the magnifying glass to focus and reveal these details, was particularly effective. 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗟𝗶𝗴𝗵𝘁 I used light as my method of interaction with the img2img diffusion. This approach felt special right away. There was something magical about holding a physical light source and seeing it influence the generated visuals. I iterated on this technique with themes like Rococo architecture, flowers, Brutalist architecture, hidden worlds, and origami landscapes. 𝗜𝗻𝗸 + 𝗛𝘆𝗯𝗿𝗶𝗱 𝗙𝗼𝗿𝗺𝗮𝘁𝘀 I also used ink in milk as a means of physical interaction with the diffusion pipeline. As I drop ink into milk, shapes come alive instantly. By learning to manipulate the combination of physical and digital elements, I steered the generated output toward my areas of interest. These iterations extended beyond ink in milk to include the format in which these elements are contained: a circular plate or a triptych of small stainless steel trays. These formats provide a structured yet flexible framework to explore themes and narratives across multiple visual spaces. It's magical. Some of that last iteration has been captured in this insightful article by Fast Company: #stablediffusion #realtime #ai

Hugues Bruyère

86,381 просмотров • 2 лет назад

Teaser trailer to my short film “Je t’aime, Julie” based on the chronicles of Julie D’aubigny who is a bisexual swordswoman in the 17th century ⚔️ Directed by Me and my brother! We are doing a festival run at the moment and using this as a proof of concept for a feature length film! Thank you all so much for the support, I can’t wait to share this project with all of you ♥️ Here is our kickstarter with more information if you are interested:
0:25

Teaser trailer to my short film “Je t’aime, Julie” based on the chronicles of Julie D’aubigny who is a bisexual swordswoman in the 17th century ⚔️ Directed by Me and my brother! We are doing a festival run at the moment and using this as a proof of concept for a feature length film! Thank you all so much for the support, I can’t wait to share this project with all of you ♥️ Here is our kickstarter with more information if you are interested:

Fiona Nova

1,776,958 просмотров • 3 лет назад

The game engine I'm cooking (RAMEN) is finally coming together! 🍜 Anyone down to test it out? I'm droppin the first version next week and it's gonna include: · Video to spritesheet generator (AI video compatible) · Spatial detection for 2D backgrounds for custom lights and shadows · Node-based game logic. Totally no-code · Tons of other stuff focused on making a 100% playable 2D adventure in just a couple of days This is just the tipo of the iceberg, so saty tuned for more details, but if you're intersted on testing it, just let me kno in the comments. Thanks!!
0:41

The game engine I'm cooking (RAMEN) is finally coming together! 🍜 Anyone down to test it out? I'm droppin the first version next week and it's gonna include: · Video to spritesheet generator (AI video compatible) · Spatial detection for 2D backgrounds for custom lights and shadows · Node-based game logic. Totally no-code · Tons of other stuff focused on making a 100% playable 2D adventure in just a couple of days This is just the tipo of the iceberg, so saty tuned for more details, but if you're intersted on testing it, just let me kno in the comments. Thanks!!

TechHalla

113,177 просмотров • 5 месяцев назад

AI can't build complex websites, they said. So I vibe coded a hyper realistic 3D tunnel gallery using Three.js with Gemini 3.1 pro in Google AI Studio in < 1 hour. It runs on Catmull Rom splines with a treadmill style sliding geometry, custom fragment shaders with depth based fog, and tight optimizations like texture disposal, disabled frustum checks, and capped DPR to keep it smooth even on long infinite scrolls. All of this is seriously hard to learn and build manually, but AI pulls it together so well! Live: Code:
0:32

AI can't build complex websites, they said. So I vibe coded a hyper realistic 3D tunnel gallery using Three.js with Gemini 3.1 pro in Google AI Studio in < 1 hour. It runs on Catmull Rom splines with a treadmill style sliding geometry, custom fragment shaders with depth based fog, and tight optimizations like texture disposal, disabled frustum checks, and capped DPR to keep it smooth even on long infinite scrolls. All of this is seriously hard to learn and build manually, but AI pulls it together so well! Live: Code:

The Bugged Dev

18,762 просмотров • 4 месяцев назад