WilliamTong retweeted
RTパストレはVRAMの制限が厳しいので、RTパストレのノウハウを活かしてCPUパストレ実装している。低クオリティのDLSSではRTほどでは無いけど悪くない。
WilliamTong retweeted
ComfyUI-Tesseraを試してみました✨
MiniMax-H3 Ref2VAを利用した動画の高解像度再生成
480×864で生成し顔が潰れている動画を3倍の1440×2592に再生成し潰れていた顔や細部が改善されるかテスト
動画1⃣ 元の動画から顔の違いを見やすくするためクロップし拡大
動画2⃣元の動画
#MiniMaxH3 #ComfyUI
🤖 Made with AI
WilliamTong retweeted
画像1枚から編集可能な3Dシーンを生成「WorldGen」🌍
🔹物体を個別メッシュ化し、選択・移動・再編集
🔹素材・照明を含めBlenderへ書き出し
🔹全てがメッシュではなく背景は3DGS。
投稿:Qixuan (QX) Zhang氏
情報元はリプ欄。
WilliamTong retweeted
Amazing stuff!
MatAnyone-E runs 4K video matting at 28 FPS on an H200 and fits on a 3090 without crashing.
You get hair strands and soft transparency without a green screen
- 7.4x speedup
- drops peak GPU memory to 7.6%
shuyun-wang.github.io/projec…
WilliamTong retweeted
【動画から高精度な法線マップを生成するComfyUIカスタムノード「ComfyUI-VizClick-NormalCrafter」】
動画を入力として、ビデオ拡散モデルベースの「NormalCrafter」モデルを活用し、サーフェス法線マップを高精度かつ時間的一貫性を持って生成できるComfyUI用カスタムノード群が公開されました!
3D再構築、リライティング、レテクスチャリング、動画編集をはじめ、H3 MinimaxやSeedanceにおける動画リファレンスとしての活用が可能です。最大の特徴は、オプティカルフローを利用して残存するちらつきを削減する「Normal temporal stabilizer」ノードや、チャンク単位のデコード機能(`decode_chunk_size`)が統合されている点ですね!これにより、低スペック環境でも高速に処理を実行できるようになっています。
#ComfyUI #NormalCrafter
WilliamTong retweeted
Little teaser of my compositing software. Nuke compatible (except for the proprietary or licensed node like Keylight, Furnace etc). OCIO, ACES, all the Blender nodes that are not in Nuke, roto, paint, cryptomatte, caching, MCP server built-in and openFX compatible too! All features are in place, now it's debugging time. It will take a while. How much would you pay for this? #compositing #nuke #fusion #b3d
WilliamTong retweeted
Your AI isn't bad at design. It has just never seen good design.
I've tested 100+ vibe coding resources this year. These are the 40 I actually kept, grouped by what you're stuck on.
Bookmark this 🔖
1. 👑 scrolltide.co
Why it's different from everything below:
- Every one is a complete prompt, not a snippet you wire up yourself
- 3 steps: copy the prompt, paste it into your AI, launch it live
- 300+ award-winning prompts, all in one place
- Full source code to download and edit
- Works with Claude Code, Cursor or v0
Built for founders and freelancers who want award-level sites without a design team or a week of tweaking. If you only open one link in this post, make it this one.
2. designmd.ai - design systems as markdown your agent can read
3. vibeprompts.dev - ready-made prompts for dashboards and landing pages
4. minimal.gallery - curated high-end sites
5. kage.design - real UI inspiration mapped to prompts
6. styles.refero.design - 2,000+ real product styles with typography
7. component.gallery - 2,600+ examples of how top design systems solve the same element
8. godly.website - curated collection of beautiful websites
9. awwwards.com - the best of web design, recognized and ranked
10. landing.love - showcase of the best animation websites
11. land-book.com - curated landing pages
12. mobbin.com - mobile design patterns
13. hoverstat.es - experimental web design projects
14. navbar.gallery - navigation patterns
15. footer.design - footer layouts
16. cta.gallery - conversion-tested forms, popups and buttons
17. supahero.io - curated hero sections
18. pricingpages.design - the best pricing page examples
19. 21st.dev - community component marketplace that plugs into agents via MCP
20. ui.shadcn.com - still the gold standard
21. ui.aceternity.com - 200+ animated React and Tailwind components
22. magicui.design - 150+ animated components and effects
23. motion-primitives.com - advanced interactions
24. uiverse.io - open-source UI elements
25. smoothui.dev - animated components built on Motion and GSAP
26. shadcnblocks.com - 2,000+ pre-built section blocks for shadcn/ui
27. kinetics.colorion.co - 150+ motion effects with React code and prompts
28. microkit.co - micro-interactions
29. glass.samasante.com - glass refraction components
30. animejs.com - lightweight animation engine
31. threejs.org - the most popular web 3D library
32. shadertoy.com - the standard online shader editor and community
33. thebookofshaders.com - free interactive intro to fragment shaders
34. threejs-journey.com - full Three.js course with shader chapters
35. kitbitz.art - 2,000+ hand-drawn illustrations
36. 3dicons.co - 3D icons
37. claude.com/product/claude-co… - command-line AI coding agent
38. cursor.com - AI-first code editor
39. v0.dev - Vercel's prompt-to-component tool
40. lovable.dev - AI app builder in the browser
Which one should I add next?
WilliamTong retweeted
LTX-2.5 🤠 IC LoRA SDR To HDR
Converts 8-bit standard-dynamic-range (SDR) video into 16-bit high-dynamic-range (HDR) video, using the SDR clip itself as the control signal and writing EXR output in addition to an HLG master.
huggingface.co/Lightricks/LT…
WilliamTong retweeted
DJI Osmo 360で2km(40分)歩いて撮影した戸隠神社奥社の参道を3D Gaussian Splattingにしました。
赤い茅葺屋根の随神門から先が有名な杉並木です。
DJI Studioで変換した全天球動画からSpirula Studioで3DGSを生成しています。(デュアル魚眼動画は上手くアライメントできなかった。)
WilliamTong retweeted
This rig in After Effects is really fun.
The setup: as the last point of the path gets closer to the first, it starts forming a circle. That same point also controls the size of the circle and where it starts along the path.
Layout to Render (Beta) drops today.
Your 3D blocking transforms into a finished shot.
Set your composition in 3D and LTX generates the surrounding environment, holding strict alignment with your CG. Shape the visual treatment using a prompt or reference image without altering your established framing.
Start from your current layout today.
WilliamTong retweeted
AMB3R-SLAM: Kilometer-scale SLAM
with Hierarchical Backend
hengyiwang.github.io/project…
TL;DR: 1) A training-free, model-agnostic SLAM system that scales to kilometer-scale trajectories in real time on one consumer GPU; 2) A hierarchical backend that enforces local, mid-level, and global consistency without bundle adjustment, so dynamic scenes work out of the box; 3) One framework that extends to stereo, RGB-D, and LiDAR.
WilliamTong retweeted
モーションデザイナーのMike Gaynor氏が開発!
タイムラインとノードグラフを融合したモーションデザインツール!パーティクルやSolver、Cryptomatteなど多数の新機能を搭載!Win&Mac&Linux
Caddis 1.1.0
3dnchu.com/archives/caddis-1…
by @CaddisApp
WilliamTong retweeted
wip
コンポジット - Aftereffects
モデリング 、アニメーション - blender
レンダリング、背景、FXなど他いろいろ - houdini
今のペースで作ってると多分11月末くらいに終わる。
WilliamTong retweeted
Modeled Earth
love how the sun light realistically gets redshifted by the Rayleigh scattering 🤩 #blender3d
WilliamTong retweeted
Minimax H3 Character Swap LoRA, I tried it locally on my RTX5090 in ComfyUI.
Swapped the character, kept the whole scene.
Links below👇
MiniMax-H3 😀↔️🤩 Character Swap LoRA
It aims to replace a selected character using an image reference while preserving the source scene
👇
huggingface.co/akatz-ai/Mini…






















