brain rotting, but the thoughts still racing // Testing img2vid lipsyncing with the InfiniteTalk model on @wavespeed_ai and a song I made in @suno v5. Love how it handles images with multiple faces and how it animates even when it doesn’t see a mouth. Quick tutorial in 🧵👇 (1/6)

🇨🇳 China’s Bytedance rolled out an AI video editor that handles video comprehension better than Gemini 3 Pro. Vidi2, a video model that can accurately find actions and objects in long videos from text. Clearly outperforming strong commercial models on new retrieval and

Nano banana pro excels at fictional scenarios. “Give me the general assembly diagram for The Device (run with it)” “A photo of The Device, taken hastily by a secret agent” “A satellite photo of The Facility, taken by The Organization & plan” “Operational status report”

I just read this paper called "Chain-of-Visual-Thought (COVT)" and it basically teaches VLMs to see and think at the same time not in text, but in continuous visual tokens. Here’s the wild part: Instead of forcing models to reason through words (which destroys all the

📢 Cascadeur 2025.3 is live! Big update with AI interpolation, Filament renderer, quadruped support, and improved physics tools. — Inbetweening is now AI interpolation — Filament (experimental, Windows): HDRI, lighting, materials — Quadrupeds (alpha): Quick Rigging +

Great tips!

🚨 DeepSeek just did something wild. They built a math model that doesn’t just solve problems, it checks its own proofs, criticizes itself, fixes the logic, and tries again until it can’t find a single flaw. That final part is the breakthrough a model that can verify its own

It’s always either a lesson or a win. Forward is the only direction I know✨ @wavespeed_ai x @Hailuo_AI

AI 3D workflows for web are getting scary good. Here is a demo with a large splat background (+LoD), mesh character, spatial audio, VR support, and lighting. I was able to create it in a few hours (repo in reply). built with @threejs @sparkjsdev @theworldlabs Audio on :)

Nanobanana Pro @higgsfield_ai なるほどこれは便利！！ いろんなショットを一発で出して、そこから選ぶ感じ 見事にアップスケールしてくれる プロンプトはリプ欄の元投稿を使わせていただきました ””” <instruction> Analyze the entire composition of the input image. Identify ALL key

🚀🚀🚀Hunyuan 3D Studio just leveled up to 1.1! We've integrated the art-grade 3D generative model, Hunyuan 3D-PolyGen 1.5, to deliver the industry's most advanced mesh quality directly to your workflow. 🎨 🖌️ Art-Grade Quad Mesh: We've pioneered an end-to-end native quad mesh

🚀 Real-time Gaussian Splat Morphing Unlike typical cross-fades, this actually moves each splat through 3D world space from source → target positions. @theworldlabs Features: -Per-splat speed control with stagger effects -3D turbulence noise for organic paths -8 transition

China's Bytedance just dropped an AI video editor that understands video better than even Gemini 3 Pro. Vidi2 can take in a bunch of footage many hours long and a prompt, and construct a script and generate a TikTok or movie from them.

FLUX.2 is now open-sourced and live in ComfyUI on Day 0！ 4MP photorealism, professional lighting/skin/fabric detail, enhanced text rendering, and 10-ref consistency. Available on Comfy Cloud & local: - FLUX.2 Dev: BF16 / FP8 partnered with @NVIDIA_AI_PC - FLUX.2 Pro API: via

Google just dropped "Attention is all you need (V2)" This paper could solve AI's biggest problem: Catastrophic forgetting. When AI models learn something new, they tend to forget what they previously learned. Humans don't work this way, and now Google Research has a solution.

Agibot A2 wins a Guinness World Record for traveling 106.3km (~66 miles) Suzhou to Shangai - using battery hot-swapping. 2026 will be the year of robotics.

Can’t stop watching the time lapse videos of “construction” from the AI interior design accounts (from inspiringdesignsnet)

🚀 Gaussian Splatting Performance Breakthrough Achieved 6M photorealistic splats rendering at 60-80 FPS in real-time in Unity. Mixamo character conversion + GPU compute shader skinning = photorealistic animated characters. Key optimizations: ✅ GPU compute shader skinning ✅

As the tools improve, I suspect we’ll see direct animation of entire mood boards. Instead of single renders, put several alts and frames in one 4k image and animate the whole thing, then select the best nodes for upscale. Fishing net the latent space for first drafts.

QwenEdit-2509 Character Turnaround Sheet LoRA: creates a composite image of a character from multiple angles, slightly improving the base model's rotation ability. https://t.co/VVIpkixzcA

🤯🍌🤯 Gemini 3.0 can convert a 2D blueprint into a 3D-rendered visualization - at 4K resolution. Just paste your 2D plan and add the prompt (in ALT). Absolutely insane for architects or interior design workflows.