thasareezy

dev @dubdotsocial

[0xMarioNawfal @RoundtableSpace]

this might be the only prompt you need for high-end motion graphics with Claude Opus 5.5.

it turns Claude into a full motion design pipeline:

•⁠ ⁠studies reference videos frame by frame
•⁠ ⁠storyboards before animating
•⁠ ⁠builds scenes in GSAP + Three.js
•⁠ ⁠renders and critiques every section
•⁠ ⁠checks motion, contrast, audio and transitions
•⁠ ⁠keeps iterating until the final video passes the quality bar

it also uses this open-source motion kit:

⁠ https://t.co/r2MwiJ8fv2 ⁠

full prompt below.

[ElevenLabs @ElevenLabs]

Introducing Eleven v4 and Eleven v4 Turbo, our fastest and most emotive voice models yet.

Ranked #1 by Artificial Analysis.

[Chris @everestchris6]

claude opus 5.5 is f*cking insane

it just one-shotted this whole video, and every frame of it is code. with the right setup you can easily sell videos like this to businesses

just wrote a whole guide on how to set this up:

[LOOPY @VibeEverything]

Opus 5.5로 만든 바이럴 영상 389개 모음

요즘 화제인 Opus 5.5로 만든 모션그래픽부터
설명 영상, 3D 장면, 게임까지 정리돼 있음

원본 영상과 리메이크 영상을 같이 볼 수 있고
프롬프트도 바로 복사해서 활용 가능

마음에 드는 영상을 내 로고, 컬러, 폰트 같은 에셋과 함께
클로드에게 주고 구성안을 뽑아달라고 하면
아이디어의 방향을 정하고 구체화하는데 도움이 됨

[dawood46 @notdwd]

STOP asking Opus 5.5 to invent motion from scratch.

good motion still starts with taste + creative direction. when you give AI nothing, it falls back to the same generic shit everyone else is posting.

left is @thednyx’s original. right is opus's version rebuilt with my own brand.

same motion, same cuts, same timing. ~15 minutes, $0.

give it great work to study, then let it handle the execution.

full prompt below ↓

<inputs>
Ask me for: my product name, a logo (or let you draw a simple mark), my brand colours (or pull them from my logo), the one-line thing a user types into the prompt box, the page that answers it (title + 2–3 sentences with one key phrase), two feature names for the stacked cards, and a music track. If I skip any, use the defaults: product "Frame by Frame" living inside its Whop hub, a viewfinder mark (four corner brackets around a bold "FF"), prompt "Make a launch video for my app", a lesson page titled "2.1 Choose a reference", cards "Launch" and "Sound", and Mixkit's free house track "Rising Forest" slowed to 124 BPM.
</inputs>

<direction>
A 12 second square product sting, 1080x1080, frame numbers at 29.97 fps (f0–f359), master rendered at 59.94 fps. Apple-keynote feel: soft, white, bright blue, glassy. The camera never cuts on a still frame: every shot enters already moving (exponential ease-out, 12–19% of the remaining distance per frame) and leaves on an accelerating move or a blur ramp. Blur follows speed and direction on every move.
Palette: my brand colours mapped onto these roles; if I give none, use page #FDFDFB, haze #B7CFEB, ice #E6F0FA, deep blue #294376 → #769CC2 sky gradient, navy #1E2F52, accent blues #2F6BFF / #3CC8F0 / mint #4ED6A0, white. Type: Inter (400/500/600/700). No purple, violet, magenta or orange anywhere.
Story: the product header rises out of a blue haze inside a light app window → a cursor glides in, turns to face where it's going, grows on hover and clicks the logo → hard cut on the music drop to the app icon with four squares orbiting into a cross → the icon collapses into a Mac menu bar → the cursor clicks the menu-bar icon, a frosted glass prompt box springs out and the prompt types → whip-tilt up through a light-blue flash into the answer page scrolling into place → a card rises over it → blur-dissolve to two stacked cards with giant frosted-glass titles → blur-dissolve to the lone logo disc → hard cut to a black end card with a glowing wordmark.
Banned: crossfades, frozen frames (except one hold in S7), stock UI kits, glows on UI text, Math.random, anything that looks like a template.
</direction>

<structure>
Coordinates are px on the 1080 stage. Music beat k = 0.048 + 0.4838·k s (124 BPM, one beat = 14.5 frames). The three hard cuts f72, f101, f159 land 2 frames before a beat: keep these exact frames.
S1 f0–71, header + window + click: haze #B7CFEB fading to #FDFDFB by f28, keep a blue floor glow #DBEEFD at the bottom. Header on one line: logo disc ⌀132 (#FBFCFE, 1px rim #E8E8EA, dark mark), the product name (Inter 500), a dark capsule pill "on Whop" (#2F2E2F, white text); the whole lockup spans x474–1063, centre line rises y932 (f0) → 776 (f2) → 683 (f5) → 608 (f10) → 546 (f24) → 536 (f37), drifting 1 px/frame after. Name appears f2 blurred, pill f4–5 heavily blurred and sliding 15 px left as it sharpens. A light hub window (radius 93, fill #FCFDFF, top band #F1F6FF, blue inner floor glow) fades in around it: opacity 0 to f14, 0.53 f15, 0.7 f19, 1.0 f24; top-left corner (408,405), runs off the right and bottom. Inside: a search pill in the top band, a left icon column (Courses icon at (537,739), faded Chat icon at (537,900)), and a course card from (764,700) off-frame with its cover image, title and "Course · 8 lessons". Cursor (black macOS arrow, white outline, 32x39) appears f27 at (891,393), glides left decelerating to (770,383) f44, rotates to point down-left f47–51 as it dives, lands on the disc's lower right (629,564) f56 → (587,546) f59 → (555,531) f71; grows ×1.55 on hover from f56; a soft ice ring (#D5F3FF → #F7FEFF, outer ⌀174) lights around the disc from f53. Camera zooms about (540,540): 1.0 f46 → 1.2 f60 ease-in-out, holds f61, then eases out accelerating to 1.04 at f71 while the cursor presses (shrinks 15% f69–71).
S2 f72–100, icon + orbit: navy squircle app icon (#294376 → #1E2F52, white mark), 276 px at f72 shrinking ease-out to 178 px by f86, radius 28% of width, on a grey halo disc #E9E9E7 growing ⌀240 (f75) → 326 (f86). Four 92 px squares (radius 26) spin in counter-clockwise, decelerating, and lock into a cross at orbit radius ≈216 by f86: white (1px #E3E8EF edge + faint shadow) left, #2F6BFF top, #3CC8F0 right, #4ED6A0 bottom. f88–100: the icon shrinks accelerating to ≈40 px, the squares slide into a row on its right (white slips behind the icon), blur ramps 0.3 → 6 px.
S3 f101–158, menu bar + prompt: white page above a black laptop bezel band (y425–475, top highlight #686866), a dark navy menu bar (y477–538), wallpaper below = the blue sky gradient with thin white line art (one big circle, two horizontal lines, one vertical, a four-point sparkle at a crossing, soft teal glow top-right). Menu bar right cluster in white: Wi-Fi, battery, toggles, the product mark at x531–584, three ⌀36 dots #2F6BFF / #3CC8F0 / #4ED6A0 at x612, 661, 709, "Mon Jun 22 9:41 AM" 34 px. Enters blurred 3 px and settling by f110. Cursor rises from below (f102), sits on the mark, presses f113–117. Camera pans content right +125 px f115–130 (fastest f118–120). A frosted glass box (white-blue glass over the sky, bright top rim, radius 60) springs out from under the mark f116: width peaks 744 at f126 and settles 726x228 by f138 around x196–922, y568–797. The prompt types from f121 (first legible "Make ") to f150 (complete) at about 1 char/frame with a 1-frame hold every 2–3 chars; caret always on; three white outline icons along the bottom; send button #2F6BFF ⌀51 with a white up arrow. From f136 the whole scene drifts up, accelerating into a whip-tilt (f158 moving ≈25 px/frame, vertical blur ≈6–8 px) while the page tints #E6F4FE over f154–158.
S4 f159–186, answer page: a light course lesson page (breadcrumb, title, body): text column x120, body 47 px Inter 400 grey #BCBCBA, line pitch 58, one key phrase ("frame for frame") in black 600. It arrives smeared and 420 px low, scrolls up with offsets 420, 315, 210, 170, 140, 116, 96, 81 (f166) … 14 (f175) … 0 (f180), then creeps −6 px by f186. Blur 24 px (f159) → 1.5 (f165) → 0. Flash #E2F4FE fading to #FDFDFB by f165. A cursor pointing straight up rides the scroll and stops under the key phrase (≈(450,591) f180), then drifts right.
S5 f187–214, card: a white card (x203–878, runs off the bottom) rises over the page: cover image 627x536 inset 24 px, radius 64, a light grabber bar at its top centre, caption semibold 38 px black, sub-caption grey 23 px. Cover top y642 (f187) → 456 (f191) → 395 (f196) → 369 (f200) → 348 (f205), then keeps drifting up ~4 px/frame. Card blur peaks 3.8 px at f189, sharp by f201; the page behind blurs to ≈2.5 px. Exit f207–214: card shrinks ~5% and rises while the whole frame blurs 1 → 7 px; cut at the blur peak.
S6 f215–244, two cards: white page, two stacked cards 538x348 (radius 57, gap 36) centred on x540, top card settling at y168 by f230, bottom at y552. Each: cover art (no text baked into it), a frosted pill top-left ("Module 3" / "Module 5"), a frosted round "•••" top-right, and a huge bold title along the bottom edge made of frosted glass (a blurred, lightened copy of the image clipped to the letters, cut off by the card's bottom edge): "Launch" and "Sound". Both enter blurred 12 px and sharp by f224; the top card enters 8% large and rises from y264; the bottom card rises from y927, staggered behind it. They drift up 3 px/frame f230–238, then accelerate up and blur out into the cut.
S7 f245–300, logo disc: page #FDFDFB, disc ⌀168 #F6F6F6 with the dark mark, rises into the centre (top y569 f245 → 491 f250 → 468 f255 → 456 f269) with a vertical smear on the cut frame, holds still f269–287 (the only frozen stretch), then shrinks accelerating to ⌀123 at f300.
S8 f301–359, end card: black radial background (#020204 corners, ≈#272729 around the word), the wordmark in Inter 600, white #F3F3F5 with a tight glow plus a wide soft halo, centred (540,540). Word width: ≈1650 px f301 (horizontally smeared, zoom streaks) → 1350 f302 → 1110 f303 → 1049 f304 → 734 f309 → 678 f311 → 563 f320 → 516 f342 (≈1 px/frame shrink) → 492 f350, then collapses ease-in: 450 f355 → 267 f359 with blur rising to 3.5 px. The film ends mid-collapse.
</structure>

<build>
1. One HTML page, 1080x1080, drawn by seek(t) as a pure function of the frame number. No CSS transitions, no timers, no Math.random (seeded hashes only). Shots register as {f0, f1, render(localFrame)}.
2. Every value is continuous in the frame number (the 59.94 master renders half frames): animate with keyframe tables kf(frame, [[f, value], ...], ease) and per-frame lookup tables with linear interpolation. No Math.floor on motion.
3. Blur: CSS filter blur for round blur, SVG feGaussianBlur with separate x/y stdDeviation for directional smears. Zoom smear on the end card = 20–30 scaled, faded copies of the word. Frosted glass = a blurred, lightened copy of what's behind, clipped to the shape.
4. Cursor: one SVG macOS arrow (black fill, white outline, soft shadow) with rotation and scale, reused in S1, S3, S4.
5. Sound (no voice), exactly 12.075 s: music at 124 BPM, soft intro, the drop at 2.47 s (beat 5, the f72 cut leads it by 2 frames). Synthesized SFX: soft impact 0.10 s; whooshes peaking at the cuts 2.402, 3.370, 5.305, 10.043 s; transition hits exactly on 6.240 s and 8.175 s; clicks at 2.33 and 3.83 s; very quiet key ticks every ~32 ms over 4.04–5.00 s; a soft shimmer at 10.05 s. Master to −14 LUFS, true peak −1 dBTP, no fade except the last 60 ms.
6. Render with Playwright (one screenshot per frame, fonts loaded first), encode H.264 yuv420p at 60000/1001, mux the audio.
</build>

<gotchas>
Measure text only after the fonts load. A long product name won't fit where a 5-letter name did: scale the whole lockup (disc gap, name, pill) to fit the span x474–1063, don't let the pill fall off-frame. A white orbit square vanishes on the white page without a 1px edge and a faint shadow. Don't put images with their own text inside the S6 cards, or the glass title doubles up. Keep the cut frames exact even where they don't sit on a beat. Nothing freezes except S7 f269–287. Heavy blur tables can wipe a shape out completely: if a frame looks empty, lower the blur until the shape still reads.
</gotchas>

<start>
Ask me for the inputs. Then show me 4 stills (f40 header in the window, f86 icon cross, f150 finished prompt, f230 the two cards) before you render the full film.
</start>

[leo @leomeethewoo]

https://t.co/V5ASL8Yn39

[Whop @whop]

Neobanks are great businesses that are easy to run.

Today, you can create your own in 15 minutes on Whop:

https://t.co/UbO5h9H5QQ

[Higgsfield AI 🧩 @higgsfield_ai]

Introducing Higgsfield x Claude Opus 5.5 for Motion Design.

With Higgsfield MCP, Opus 5.5:

&gt; Generates images, logos and graphics
&gt; Builds and animates compositions
&gt; Renders the final video

You get an editable After Effects project to keep working on.

[Himanshu Hingorani @himanshubuildss]

I've tested every vibe coding resource I could find this year.
These are the 28 I actually kept, grouped by what you're stuck on.

Bookmark this one 🔖 🧵👇

1. https://t.co/RlChWqfItA - 200+full build prompts for 3D and scroll-driven sites, each one a complete design brief rather than a component you drop in.
2. Minimal Gallery (https://t.co/iRLY9taJAo) — curated high-end sites to feed your agent
3. Kage (https://t.co/hOf1AxdD3z) — real UI inspiration mapped straight to prompts
4. Refero Styles (https://t.co/bhB2GfGasZ) — 2,000+ real product styles with typography
5. Component Gallery (https://t.co/ZNFmqbq8CR) — 2,600+ examples of how top design systems solve the same element
6. AppShot Gallery (https://t.co/O2mKE7krky) — real app screenshots for mobile
7. Navbar Gallery (https://t.co/75zV2io5XW)
8. Footer Design (https://t.co/XPn2VgZ3ob)
9. CTA Gallery (https://t.co/NDaj9DSJrG) — conversion-tested forms, popups, buttons
10. 404s (https://t.co/daJ9Ub8h0U)
11. DESIGNmd (https://t.co/zii2nKAcsS) — design systems as markdown your agent can read
12 VibePrompt (https://t.co/mHD5eSDhVq) — ready-made prompts for dashboards and landing pages
13. https://t.co/p6pXo8h81P — component registry that plugs into agents via MCP
14. Kinetics (https://t.co/NvM0WblvCf) — 150+ motion effects with React code and prompts
15. shadcn/ui (https://t.co/vUKwdPJtTf) — still the gold standard
16. Aceternity UI (https://t.co/nxASYyV9W9) — 200+ animated React/Tailwind
17. Magic UI (https://t.co/E2oLbxc3Zv)
18. Motion Primitives (https://t.co/MIhQHgyn0c)
19. Uiverse (https://t.co/9G9JFLtfSI)
20. UIAble (https://t.co/E5NS5shItY)
21. mapcn (https://t.co/AckvMvEcod) — map components, markers, routes, popups
22. MicroKit UI (https://t.co/iFJCleZrAb) — micro-interactions
23. Liquid Glass (https://t.co/cP8YFdazmT) — glass refraction components
24. CSS Text Effects (https://t.co/yHgo53XFqY)
25. Circle Loaders (https://t.co/hUZrcbbaBX)
26. Gradient Buttons (https://t.co/qYOdx7Bj1s)
27. Kitbitz (https://t.co/L7sX9ESnU6) — 2,000+ hand-drawn illustrations
28. 3Dicons (https://t.co/BfrrSJOj6q)
29. Anime.js (https://t.co/knFchywxRi)

[leo @leomeethewoo]

If you want to make product videos like Apple

Bookmark the article below

[Deedy @deedydas]

You can now generate an entire 3blue1brown style video from any research paper with Opus 5.5.

Here’s a 8min video summary of “Regularized Recursive Self Improvement of Agent Harnesses”.

The 90%ile educational YouTuber is fully automated.

[donald @donaldjewkes]

I made this with one prompt using Opus 5.5

I spoke to my computer for 5mins, claude worked for 12 hours, and I woke up to this

full prompt:

[Addy Osmani @addyosmani]

Claude Opus 5.5 drew each frame of this animation in JavaScript. Here's "how browsers work" in 40 seconds.

[JP @jpthor]

Here we goooooo

Remember I said we needed a 10x on 3 fronts?

1) intelligence
2) bandwidth
3) latency

This is (3) and it has been achieved.

[stdrc @istdrc]

TL;DR: We just made Raft source-available.

We don't accept pull requests. We want prompt requests instead.

We are exploring what a new contribution model for open source could look like in the age of agents.

[雨夹雪❄️ @mizorewww]

介绍比Jev快50倍，在你设备上跑的laya-mlx!

只在你的设备上占用最高1G内存

Laya是一个开源的类似于Jev的,基于文本输出概率的分类系统

我将其移植到MLX,并且做了一些性能优化!

视频中就是这个模型在我的本地M3Max上玩贪吃蛇

这个模型能够以每秒决策60次的速度玩贪吃蛇！

https://t.co/31KGUiNunb

[Ashutosh Purohit @ashutoshpuro97]

I think I’ve found a pretty interesting use case for Jev (@typesafeai).

An always-on AI that knows when you’re talking to it and when you’re speaking to someone else

A thing that frustrates me about Siri/Alexa and even ChatGPT live is that there are moments that make experience feel unnatural.

With Siri/Alexa you keep saying a wake word to get their attention and then talk

With ChatGPT voice- it’s a 1:1 conversation. If someone sitting next to you says something and you reply to them, ChatGPT doesn’t understand you were talking to them.

But what if your agent could figure that out?

So here's a small demo called Toothless - where Jev figures out in real time whether something you’re saying is directed at the AI or someone else.

It does this by considering multiple aspects simultaneously - like whether this is a continuation of a topic that was discussed before, whether it’s something that uses the internet or something can be helped by AI agent and a couple of other things.

I think with a low cost classifier we are much closer to being able to have an always on ambient AI, which can jump and help us naturally, without requiring us to press buttons or keep calling out a command.

Exciting times ahead!

[brain function collapse @brainFnCl]

This is NOT Jev.

Open source. Runs on your laptop. Decides in ~27 ms, about 200× faster than waiting on a hosted LLM.

Here it is playing Tetris by itself 👇
https://t.co/eq4gP53o2A

[Marcus Lowe @marcus_lowe]

what if copy/paste was smart?

powered by @typesafeai jev

it feels like every computer interaction will get rewritten

[Guillermo Rauch @rauchg]

Jev Generative UI has been achieved externally