OpenX retweeted
Ask an AI to animate a character and it looks like a slideshow: the face teleports between poses, everything stops dead, blinks flicker.
This free skill teaches your coding agent how animators actually do it: one clock for the scene, parts that turn on their real pivots, hair and cloth that follow a beat later, idle breathing that never stops.
It works on characters, creatures, plants, fire, water, logos, icons, charts and whole scenes, at 60 fps from any SVG.
Install:
npx skills@latest add rossark0/rossark-motion-skill
OpenX retweeted
Qwen3.5-4B-3D-Coder w/ Three.js
A 4B parameter model can generate interactive 3D scenes using Three.js.
Qwen3.5-4B-3D-Coder's October 29 update makes the full FP16 weights, training dataset, and technical report publicly available.
The dataset contains 4,146 synthetic examples.
Training took about 60 hours, combining full fine tuning with a short LoRA stage.
3D scenes are generated as HTML and JavaScript through Chat mode, not agent tool integration.
Specific generation settings and a Three.js system prompt are required.
Q4_K_M and Q8_0 GGUF versions are also available, with more quantized variants planned.
huggingface.co/ccore/Qwen3.5…
OpenX retweeted
Someone needs to stop Claude Opus 5.5. It turned one picture into a robot octopus I can take apart piece by piece.
i just said "make the model" and what came back floored me.
the build:
> the picture set the look, every petal, lens and link is three.js code
> opus rendered one still next to the picture and fixed the gaps for 3 rounds before it touched the film
> the first pass looked like a toy, so it went back and packed 202 links and 480 wheels into 8 arms
> every arm is one draw call, posed every frame by walking its chain of bends
> something opens every 2-3 seconds: the lens unscrews, the mantle hinges, the pump rises out spinning
> at 0:16 the whole thing blows apart on a 330° turntable and reassembles before the turn ends
25 seconds, 2,616 lines of code and one octopus that leaves the frame on its own.
the skill, the prompt and the live sim 👇
OpenX retweeted
GitHub tiene escondida una auténtica mina de oro para los que hacen cine con IA.
10 repos open source para generar vídeos, animar fotos, montar escenas de cine y automatizar la edición.
Algunos son una barbaridad.
Guárdatelo:
1. LTX-2
Genera vídeo con el audio ya sincronizado. Trae herramientas para ejecutar el modelo y entrenar tus propios LoRAs para darle tu estilo.
github.com/Lightricks/LTX…
2. ComfyUI WanVideo Wrapper
Monta flujos visuales dentro de ComfyUI para crear vídeo, animar fotos y controlar la cámara con los modelos Wan.
github.com/kijai/ComfyUI-…
3. CogVideoX
Convierte un prompt o una imagen de referencia en un clip de vídeo. Muy útil para probar escenas e ideas visuales.
github.com/zai-org/CogVid…
4. VACE
Crea y edita vídeos a partir de imágenes de referencia, metraje que ya tengas y máscaras. Puedes combinar controles para decidir qué cambia en cada plano.
github.com/ali-vilab/VACE
5. DiffSynth-Studio
Todo en uno para ejecutar modelos de vídeo, animar imágenes, jugar con los movimientos de cámara y entrenar modelos.
github.com/modelscope/Dif…
6. MuseTalk
Sincroniza el movimiento de los labios de una cara con tu audio. Perfecto para doblajes, diálogos y personajes que hablan.
github.com/TMElyralab/Mus…
7. OpenChatCut
Editas hablando con él, con una línea de tiempo de varias pistas. Puedes conectar agentes de IA con skills y MCP, y luego renderizar con Remotion.
github.com/0xsline/OpenCh…
8. FireRed-OpenStoryline
Le cuentas qué edición quieres con tus palabras y su agente de IA planifica los pasos y coordina las herramientas. Además guarda skills para que el estilo sea siempre el mismo.
github.com/FireRedTeam/Fi…
9. AutoSubs
Saca los subtítulos en local y los pasa directamente a DaVinci Resolve, Premiere Pro o After Effects.
github.com/tmoroney/auto-…
10. Editly
Automatiza el montaje con instrucciones que puedes reutilizar. Junta clips, imágenes, títulos, transiciones y música desde la terminal o por API.
github.com/mifi/editly
¿Cuál vas a probar primero?
OpenX retweeted
各位，AI 剪视频这条赛道已经卷到我看不懂的地步了。
刚刷到一个开源项目，GitHub 上直接冲到 4.2 万星。它不是帮你剪，而是压根不用你动手。
你丢一句大白话进去，脚本、素材、配音、字幕、成片，全都替你跑完。
最夸张的是，你扔一条爆款视频进去，它能把人家的脚本结构、镜头节奏、画面风格全拆解开，然后给你出好几套复刻方案。
底层 12 条流水线、100 多个工具、700 多个 Agent 技能包。这哪还算工具，简直等于直接塞给你一个制作团队。
我昨天试了一下，第一反应是——那我学剪辑图啥？说实话真有点被吓到。
链接我放这儿：github.com/calesthio/OpenMon…，你们自己看。
反正我已经丢进收藏夹了，用不用另说。但做短视频的兄弟们，这玩意儿你们真得盯一下，不然明年可能真不用自己动手了。
OpenX retweeted
Editing a talking video means hours of cutting retakes, dead air and false starts. Claude Code can now do the first pass inside DaVinci Resolve.
You film and make the taste calls. Claude:
→ cuts retakes, silences and false starts
→ keeps and places your B-roll
→ fixes and normalizes your voice
→ adds music with ducking under speech
→ writes clean captions
→ adds brand motion graphics timed to your words
Tell it "too choppy" or "music louder" and it saves the lesson, so every video comes out closer to your style.
Install:
git clone github.com/appariciojunior/v…
cp -R video-editor-skills/skills/* ~/.claude/skills/
Then tell Claude Code "set up my video editor".
OpenX retweeted
🔴 Esto es lo más peligroso de internet hoy
Alguien ha entrenado una IA que lee velas como ChatGPT lee texto.
Y la ha soltado gratis.
Se llama Kronos. Más de 40.000 estrellas.
Convierte cada vela en tokens y predice la siguiente.
Entrenada con más de 12.000 millones de velas de 45 exchanges.
No es magia ni garantiza nada. Es un modelo de investigación.
Pero puedes probarlo en tu portátil.
OpenX retweeted
10 GITHUB REPOS FOR BUILDING AI THAT CAN TURN IDEAS INTO SOFTWARE:
1. OpenHands
github.com/OpenHands/OpenHan…
2. Aider
github.com/Aider-AI/aider
3. Cline
github.com/cline/cline
4. Roo Code
github.com/RooCodeInc/Roo-Co…
5. Continue
github.com/continuedev/conti…
6. SWE-agent
github.com/SWE-agent/SWE-age…
7. Goose
github.com/block/goose
8. GPT Pilot
github.com/Pythagora-io/gpt-…
9. Plandex
github.com/plandex-ai/plande…
10. Open Interpreter
github.com/openinterpreter/o…
Save this stack, you can use it every day
OpenX retweeted
🤯 彻底疯了！本地大模型直接变天！
Unsloth 刚刚放出 DeepSeek-V4.1-Flash 最新 Dynamic V3.1 量化版，现在只要 64GB 内存就能原地起飞！ 🚀
以前跑这种体量的怪兽还得堆昂贵服务器，这次直接把门槛砸穿：
🔥 Engram 直接放 SSD：按需极速读取，硬生生把内存占用砍到大降！
⚡ 推理最高狂飙 3 倍：配合 DSpark 加速，本地跑出飞一般的丝滑感！
🏆 基准测试全面屠榜：Dynamic V3.1 在同级量化里几乎全线霸榜第一！
📚 硬刚 1M（100万）超长上下文：整个代码库、几十本书一次性直接喂进去！
64GB 内存设备直接封神，手里有机器的赶紧冲！👇
🔗 模型：huggingface.co/unsloth/DeepS…
📖 指南：unsloth.ai/docs/models/deeps…
OpenX retweeted
Uzu is an open source inference engine that lets you run AI models directly inside your iOS and Mac apps.
- Zero latency, full data privacy, no inference costs
- Simple high-level API in Python, Swift and TypeScript
- Built to use unified memory on Apple devices
- Easy to add support for new models
- MIT license
NO API BILL. NO SERVER. NO DATA LEAVING THE DEVICE. IS THIS HOW EVERY APP SHIPS AI SOON?
github.com/trymirai/uzu
OpenX retweeted
You never remember the filename. You remember what was in it.
Kensaku searches your whole computer the way you remember things: text inside PDFs and Word docs, words inside screenshots, and plain-language searches like "the PDF about database indexing" or "screenshots from last week."
It finds 100,000 filenames in about a millisecond, uses around 12 MB of memory, and nothing ever leaves your machine. Your AI agents can search it too.
Free and open source.
github.com/yjid/kensaku
OpenX retweeted
Someone is putting the law of every country into one git history.
Git for Global Law turns each country's laws into a repository where every commit is a real law taking effect. So you can see exactly what the law said on any date, and which law changed every word.
It started with the US Code, which hit #1 on Hacker News. The global version is growing country by country: Germany's constitution since 1949, Estonia and more so far.
Free to use for anything, and perfect for feeding to an AI.
github.com/lexifina/global-l…
OpenX retweeted
Latest update of PSeer AI has been running smoothly over 100 hours! It took me two weeks and I’m glad to finally deliver on the utility I promised. This is a major utility deliver for Pippy Seer.
1/4 Check the "Opportunity" for projects that are still open for participation
OpenX retweeted
5,500,000 HP. 🏮
That's Nurarihyon. 72 hours. He heals 550,000 every midnight.
5 fights a day each.
If 9,000 of you show up every day, it's close.
If fewer, he walks.
Link your X now, or you can only watch.
Tonight, 00:00 UTC.
sakurafields.com
Nurarihyon, master of the Night Parade, comes down our road on 30 October 🏮
One boss. One health bar for the whole server. 72 hours.
· 5 fights a day each. Turnout wins, not grinding.
· Every midnight he heals 10% in the dark.
· Lose a fight, you still hurt him. Win it, 50% more.
· Only X-linked accounts fight. One X, one fighter.
If he falls, every fighter gets 5,000 Petals, a Moon Scroll and a card that exists only for this night.
If he walks away, nothing. And that card is never minted.
sakurafields.com
OpenX retweeted
6 new badges just landed in the XREIGN gallery 🎖
Lucky Seat · Exhibitor · Bounty Hunter · Full Album · Hall of Fame · Grand Tour
Each one is won at the UNITS Season fair. Get paid by all four $REIGN events and Grand Tour is yours.
xreign.app/units-season
UNITS Season is live 🤖
2 weeks · 5 ways to win · 1.4M $REIGN + real UNITS
🎰 Grand Draw
Every 10 passports sealed make a round.
One of the 10, at random, wins a UNIT, sent straight to their wallet.
🖼 Exhibition
A new theme every week. Week 1: Gold Chain.
Hold one + a passport → split 100,000 $REIGN
📜 Wanted
3 traits on the posters every week.
Hold one at the snapshot → up to 5,000 $REIGN per robot
📒 Collector's Album
4 pages of traits to collect.
Complete a page → share its pool. 400,000 $REIGN in total
🏆 Hall of Collectors
Rarer robots score more.
Top 10 split 515,000 $REIGN
Passport = ×2 on Wanted, Album and Hall.
Your live ticket + the fair map 👇
xreign.app/units-season
OpenX retweeted
Most NFT raffles: 1 in 10,000.
UNITS Season: 1 in 10.
Every 10 passports sealed = one round.
One random seat in that round wins a UNIT.
No lucky number to wait for.
No whales buying extra tickets.
Same odds for every seat.
Seal yours 👇
UNITS Season is live 🤖
2 weeks · 5 ways to win · 1.4M $REIGN + real UNITS
🎰 Grand Draw
Every 10 passports sealed make a round.
One of the 10, at random, wins a UNIT, sent straight to their wallet.
🖼 Exhibition
A new theme every week. Week 1: Gold Chain.
Hold one + a passport → split 100,000 $REIGN
📜 Wanted
3 traits on the posters every week.
Hold one at the snapshot → up to 5,000 $REIGN per robot
📒 Collector's Album
4 pages of traits to collect.
Complete a page → share its pool. 400,000 $REIGN in total
🏆 Hall of Collectors
Rarer robots score more.
Top 10 split 515,000 $REIGN
Passport = ×2 on Wanted, Album and Hall.
Your live ticket + the fair map 👇
xreign.app/units-season
OpenX retweeted
Halloween is coming to the UNITS fair 🎃
On Oct 31 the lights go out, the screens die,
and the fair turns into its haunted twin.
4 doors. 50 UNITS behind them.
Tasks, friends, passports, haunted robots —
each one opens a door.
Two days left in daylight.
OpenX retweeted
HOLY SHIT. Your AI agent can do a lot more when it has access to the real internet.
Here are 10 open-source tools that help AI agents search the web, browse websites, and collect real-world data:
1. Web Scout MCP
Connect AI assistants to web discovery workflows.
github.com/pinkpixel-dev/web…
2. MiniMax Search
Find relevant information online.
github.com/MiniMax-AI/minima…
3. WET
Work with web data.
github.com/n24q02m/wet
4. Firecrawl Web Agent
Perform web research and agent-driven browsing.
github.com/firecrawl/web-age…
5. QZ Browser Agent
Automate browser interactions for AI tasks.
github.com/alibaba/qz-browse…
6. Master Fetch
Fetch content from online sources.
github.com/dondai1234/master…
7. Open WebSearch
Add open-source web search to AI workflows.
github.com/Aas-ee/open-webSe…
8. Kagi MCP
Connect AI workflows to Kagi Search.
github.com/kagisearch/kagimc…
9. Web Search MCP
Give AI assistants a web search tool.
github.com/itme-brain/web-se…
10. Oxylabs Web API MCP
Access web search and scraped data.
github.com/oxylabs/web-api-m…
The goal isn't just to make AI generate answers.
It's to help agents find information, explore websites, and work with real-world data.
Save this list for your next AI project. 🔖
OpenX retweeted
China has killed the entire vector database industry.
They open-sourced TencentDB Agent Memory. It gives any AI agent long-term memory that runs 100% locally.
No Pinecone. No cloud APIs. No repeating yourself every session.
- 61% fewer tokens
- PersonaMem accuracy: 48% → 76%
- Zero external API dependencies
- Runs on plain SQLite
Most memory systems compress your history into an opaque vector pile. when recall goes wrong, you're guessing. this one doesn't compress, it builds a semantic pyramid.
L0 Conversation → L1 Atom → L2 Scenario → L3 Persona.
Short-term state gets encoded as a Mermaid graph in your agent's context. verbose tool logs get offloaded to disk. when the agent needs proof, it drills back via node_id to the exact raw log.
no lossy compression. every layer is readable markdown you can just open and inspect.
5.1k stars. 100% Open Source.
---
I also create content around the latest AI trends and GenAI tutorials on YouTube. Check them out here: arindamm.dev/yt
