David retweeted
让微信小程序直接在普通浏览器中运行，从而使用 Chrome DevTools 调试并支持 Playwright 无头 E2E 测试。
github.com/open-wxa/wxa-runn…
David retweeted
Fibo 2.0 from @bria_ai_ is the #1 open weights model on AA-Image-Editing v2.1 and statistically tied with Qwen-Image-2.1 for the lead on AA-Image-T2I v2.1, a large step up from Bria's first-generation Fibo models on both leaderboards
Fibo 2.0 is Bria’s new 8B parameter open weights image model. The weights are released under a non-commercial license, which requires a separate agreement with Bria for commercial use. Bria released two sets of open weights: Fibo 2.0 Dev for Text to Image, and Fibo 2.0 Edit Turbo, a faster distilled model, for Image Editing. We benchmarked Fibo 2.0 through the Bria API at its default settings.
Fibo 2.0 is level with Qwen-Image-2.1 as the leading open weights model on both AA-Image-Editing v2.1 and AA-Image-T2I v2.1. It is a large step up from Bria's first-generation Fibo models on both leaderboards: it scores higher on every editing action and Text to Image capability we measure, with the biggest gains in Text or Symbol Edits, Text Rendering and Layout.
Fibo 2.0 costs $32 per 1k images on the Bria API. Luma UNI 1 Max ($100 per 1k images) and Qwen Image 2.0 Pro ($75) cost more and score slightly lower on both leaderboards, while Grok Imagine Image Quality scores about the same at $50. It is available through the Bria API and fal, with open weights on Hugging Face.
Congratulations to @bria_ai_ on the release!
See below for our analysis and example outputs of Fibo 2.0 on AA-Image-Editing v2.1 and AA-Image-T2I v2.1 🧵
David retweeted
Someone built a Claude Code for electronics 🤯
It’s called Blueprint. Describe your Arduino or Raspberry Pi project and get:
- Wiring diagrams
- A list of parts
- Step-by-step assembly instructions
From an idea to a build you can follow.
100% free.
David retweeted
Video 2 GLSL
I gave Claude Opus 5.5 a 20-second video clip. It rebuilt the whole thing as one GLSL fragment shader: 156 KB of code, no textures, no video frames. Just math, looping seamlessly
David retweeted
人物质感是可以调节得很真实的。
Kling O3 生成，分镜首帧图做了人物质感调节，
再结合提示词，可以做到几乎于真人的效果
Prompt：
画幅与景别：16:9，近景/中景（2位女孩置于画面中央至中右侧）
氛围风格：明亮自然光、阳光透过桃树叶的斑驳光影、极质朴的古风麻布衣裳、充满生命力的纪实烟火气
0–2秒：
少女们处于近景中景状态，身穿粗布素衣，站在果实累累的野生桃树下。左侧女孩微微抬手拉低带有绿叶的桃枝，眼眸清亮，视线从树梢转落到近前一颗饱满粉红的野桃上。阳光透过叶隙在她脸上形成柔和光斑。微风拂过发丝与树叶，近处桃叶轻微晃动。镜头呈现真实手持跟拍呼吸感，轻微上下浮动，带一点自然的重心漂移。
2–5秒：
右侧女孩凑近，两人眼神交汇，瞬间露出爽朗又娇俏的浅笑，眼角挤出自然的浅浅笑纹。左侧女孩动作细腻地摘下野桃，娴熟又自然的将野桃往胸前的衣服上蹭了蹭，递给旁边的姐妹，语气温柔开口说道：“今年的桃可甜了，你吃吃看” 嘴角带着宠溺的浅笑，眼神温柔地注视着旁边的女孩。嘴唇开合自然，随着说话的节奏，浓密的睫毛轻轻眨动，头部向右侧微微倾斜，呈现出极其亲昵放松的交谈状态，面部肌肉舒展。
5-8秒
右侧的女孩轻轻抬起手接过野桃，眼神带着轻松的期待，随后将桃子缓缓送到嘴边，轻轻咬下一口。咬下的瞬间野桃发出清脆的响声，女孩咬桃的动作自然停顿半秒，眉眼产生细微变化，像是突然感受到桃子清甜多汁的味道。她的眼睛微微睁亮，随后眼尾缓缓弯起，原本平静的神情逐渐变成惊喜和满足，眉梢轻轻上扬，嘴角不自觉地向上翘起。她轻轻咀嚼，脸颊随着咀嚼动作产生细微自然的变化，眼神明亮而有光。语气满是惊喜的说道：“姐姐，这野桃真的好甜呀！”她忍不住露出一个爽朗又满足的笑容，嘴角完全舒展开来，眼睛弯成柔和的弧线，脸颊微微鼓起，神态带着发现桃子特别甜之后的惊喜感。
8-10秒
她们肩膀随咯咯笑声微微耸动，笑容鲜活灵动。镜头带有极轻微的呼吸式微推，保持近景范围，不拉近到面部变形，维持自然的空间层次。
负面提示词：人物肢体扭曲穿模、手指畸形错乱，面部僵硬死板、模板化假笑，过度磨皮产生塑料假面，五官错位、睫毛僵硬杂乱；衣料静止不动、发丝没有浮动、对白生硬刻板如同背诵台词；镜头剧烈抖动眩晕、生硬旋转硬切；画面出现字幕、水印、logo、bgm、现代服饰配饰；光影灰暗扁平、画面模糊失焦、呆板居中构图、廉价 CG 质感，人物眨眼神态僵硬迟钝。
David retweeted
Prompt to 3D character animation, now with high fidelity hand + finger tracking✌️
FLUX 3 Max video → WiLoR for hands + GVHMR for body tracking → any animated character!
open source: github.com/blendi-remade/pro…
Prompt → animation at realtime speed 🕺
FLUX 3 Max generates the video, GVHMR pulls clean 3D motion from it on one GPU, and it's retargeted live onto any mixamo character. Fully automated flow!
Open source: github.com/blendi-remade/pro…
David retweeted
Real-time voice interaction is not just ASR + LLM + TTS glued together.
Human conversation is messy: we interrupt, change our minds mid-sentence, and decide when to speak or wait.
Qwen-Audio-3.1-Realtime is built exactly for that.
🤖 Made with AI
David retweeted
I made a skill that turns your app into a demo in HTML.
So they stay sharp at any size, loop, scrub, jump to any step and stand still for reduced motion.
I open-sourced it as a skill:
fil.design/skills/html-demo
David retweeted
As an AI hardware founder in Shenzhen, I sift through tons of smart devices daily. Most of them are pure noise.
So I spent the last 2 months putting together almost every single AI hardware project on the market that is actually worth seeing.
Click into any product to get its official website + GitHub repo.
Enjoy! 👇
David retweeted
Flint is the best-performing model in its class (MoE models under 50B) on agentic coding benchmarks like SWE-bench Pro v2, Terminal-Bench 2.1, and SWE-bench Verified.
It also rivals similarly sized dense models that are much slower to run.
David retweeted
Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine.
• Best-in-class fast local coding
• Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark
• Built from scratch in under 4 months
Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.
David retweeted
Introducing Topview T1 🎬
Topview’s newest AI video model. Built to rival Seedance 2.5 1080P quality at 62% lower cost.
More shots. More iterations. A much smaller bill.
Plus, Ultra Annual users get 30 days of unlimited T1 generation.
Put T1 to the test, now on Topview.
#TopviewAI #TopviewT1 #AIVideo #AIContentCreation #AIFilmmaking
David retweeted
You can now build 3D games with fluid, expressive character animation—without learning to animate each action yourself.
Manus handles the rigging and animation for you, so you can build more of your game on your own.
What used to take a specialist hours of work is now built in by default.
David retweeted
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
David retweeted
LayerFS v1.0.0 is out 🎉 Open-source (MIT)
Recursive filesystem for AI agents.
6 core ideas:
1. Ephemeral computing: agents spend ~80% thinking, so each Bash call mounts a Workspace, then commits
2. Decentralized agents: every agent commits to one Store, no DB server
3. Multiagent coordination: every tool call is a Commit; same-file clashes refused
4. Recursive filesystem: forks copy no files, at any depth
5. Ultra storage: dedup, then compression: 10 copies → 0.31
6. Instant bootstrap: ignored files stay in every Commit. ~5 min install vs 0.05 s mount
And 4 more:
7. Auto checkpoints: fork from any step
8. Append-only objects: content-addressed, never rewritten
9. .lww: lockfiles and caches take the newest edit
10. Big file, small edit: a write keeps only its bytes
Here are our links:
github: github.com/Ephemeral-AI-Lab/…
website: layerfs.ai/
Watch the video below, and have Fun!!!