Vigo Zhao

@itsshara_ai Yes, if this pipeline is viable, it could likely reduce costs by 50% to 60%.

@VigoCreativeAI If Qwen holds up that pipeline gets a tier cheaper fast.

Qwen3.8 announced: 2.4T params, open weights. As a Qwen Dev Ambassador, three things I actually care about: 1. Can it take over the meta-prompting stage of my pipeline? That slot is Opus today — if 3.8 holds up, my entire visual workflow gets a tier cheaper. 2. JSON output stability. My whole "fixed structure × variable elements" method rides on this. 3. Long-form copy quality — including Chinese, which Western evals almost never test. Max-Preview is live on Token Plan. Testing first. Full report soon. #Qwen

@mingli0x @YouArtStudio Good grief, whenever Ryu frowns, the whole world turns into equations 😂 That "he had it all calculated" trope is so cool.

You've met the analyst. Now watch him think. 🔎 The moment Ryu focuses, reality unfolds into data formulas, trajectories, and molecular structures floating in the air. He already knows how this ends. He calculated it. Ryu, in motion. Video: @YouArtStudio #FLOWBOUND #AIvideo

Which dossier should become a full series next? 1. Deep Bloom 2. Vector Run 3. Moon Trace 4. Phase Shift

This system keeps the archive grid, technical diagrams, chrome ribbons, print texture and palette logic consistent. Change the subject, mission, title, specimen and data modules to rebuild the entire series. Full JSON ↓ { "style_name": "Retro Future Chrome Portrait Dossier", "style_slug": "retro-future-chrome-portrait-dossier", "style_version": "2.1.0", "style_summary": "A retro-futurist editorial portrait system built from an asymmetric technical dossier sidebar, an edge-cropped posterized face, liquid-chrome interruptions, modular grid lines, optical diagrams, and coarse halftone print grain in midnight indigo, warm cream, hot magenta, cobalt, and coral.", "environment_variables": { "SUBJECT": "The new portrait subject or character archetype", "SUBJECT_ACTION": "The subtle action, gaze, or expressive beat", "PRODUCT_OR_PROP": "The new prop, instrument, specimen, tool, or object integrated into the portrait", "LOCATION": "The abstract editorial setting or implied environment", "BACKGROUND_ELEMENTS": "Technical diagrams, optical motifs, maps, grids, or data-panel details", "MAIN_TEXT": "A short original italic display headline", "SECONDARY_TEXT": "Original compact dossier labels and supporting microcopy", "ACCENT_SYMBOL": "A simple original star, crosshair, spark, orbit, or locator symbol", "WARDROBE_STYLE": "The new wardrobe, makeup, accessories, or styling direction", "STYLE_FIDELITY_ANCHORS": "Observable visual traits that must remain visible", "SOURCE_CONTENT_TO_AVOID": "Literal source content and identity marks that generated samples must not recreate", "ASPECT_RATIO": "4:5 or 5:4" },

Four worlds. One archive. Built from one reusable JSON system. Deep Bloom. Vector Run. Moon Trace. Phase Shift. Full reusable JSON in the first reply ↓

@eeelsi Open source makes it accessible to more people and brings the benefits of AI development to everyone

@VigoCreativeAI We are very much excited about the Chinese models in US as well, especially on how the Gov lately is taking over who gets to play with the new models and who is not.

@itsshara_ai Yes, the content structure of JSON is still

@VigoCreativeAI Four time standard still feels like the number to beat.

2026 Driver Dossier 01/04 Max Verstappen — Four-Time Standard. Dutch Attack. RB22: Built to win. Driven to the limit. Relentless in every lap. In 2026, is Max still the driver everyone has to beat? Or has the rest of the field finally closed the gap? Who’s your pick to take the championship this year? 👇 #F1 #Formula1 #RedBull #Verstappen #F1Concept #DutchAttack

@apex_led Sure thing, bro

@VigoCreativeAI Follow back fam, love the energy behind this project 🚀 interested in connecting, send me a DM 😊

@getpy Yes, but the theme of this poster should highlight sports and the figures, so the font should be relatively understated.

@VigoCreativeAI Love the minimalism, would tweak font size for readability. Lovely.

Same system. Four different sports. One design language locked across all posters: – Extreme negative space – High-contrast motion lines – Single accent color per study – Consistent typography and hierarchy Sprint Study / Corner Study / Baseline Study / Footwork Study Different athletes. Different sports. Same visual rules. This is what a reusable design system looks like when applied with discipline. Full JSON in the replies 👇 #DesignSystem #PosterDesign #SportsDesign #AIDesign #VisualSystem

I really buy into the concept, but the real test is the question MapleShaw asked: are the scenes we've walked through cached replays, or are they recalculated every time? If it's caching, then it's essentially just "smarter video"; if it can maintain consistency while calculating frame-by-frame in real-time, then it deserves to be called a world model. The fact that it's open-sourced under Apache 2.0 allows everyone to verify it in the repository themselves, which is a huge plus.

1/ 过去视频生成了一条大家卷解析、时间长、运动一致性的路线——所有指标实际上都在回答一致的问题：也许视频不能骗过眼睛。但看完Alaya World 的实机录屏，我突然觉得这个问题问错了。它并不是想生成一段更长、更清晰的视频。它压根无意生成“视频”。

Links： qwencloud.com/pricing/token-…

@Alibaba_Qwen Massive milestone. 2.4T params AND open-weight — this is the Qwen spirit 🔥 As an ambassador I jumped straight into Qwen3.8-Max-Preview on Token Plan, and it more than lives up to the specs. Huge win for the open-source community. Congrats to the whole team! 🚀 #Qwen #Qwen38

Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! 🚀  Token Plan international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…

Here’s the reusable system used for all 4 posters: Style: Minimal Grain Speedline Sports Poster { "style_name": "Minimal Grain Speedline Sports Poster", "style_slug": "minimal-grain-speedline-sports-poster", "style_version": "2.1.0", "style_summary": "A sparse sports editorial poster system built from an expansive warm-white grain field, one low-positioned high-contrast action photograph, long horizontal speed bands, tiny widely tracked uppercase typography, and a single controlled accent color.", "environment_variables": { "SUBJECT": "The main athlete or motion-focused subject", "SUBJECT_ACTION": "The decisive athletic action or movement", "PRODUCT_OR_PROP": "The unbranded sport prop or piece of equipment", "LOCATION": "The minimal sporting environment or surface", "BACKGROUND_ELEMENTS": "Sparse environmental marks and horizontal graphic structures", "MAIN_TEXT": "Short widely tracked uppercase title", "SECONDARY_TEXT": "Compact supporting phrase or event line", "ACCENT_SYMBOL": "Small separator, index mark, number, or geometric symbol", "WARDROBE_STYLE": "Unbranded athletic wardrobe and protective gear", "ASPECT_RATIO": "4:5 or 5:4" }, "prompt_template": "Create a {ASPECT_RATIO} image in the \"minimal-grain-speedline-sports-poster\" visual style.\n\nSubject: {SUBJECT}\nAction: {SUBJECT_ACTION}\nProp/Product: {PRODUCT_OR_PROP}\nLocation: {LOCATION}\nBackground: {BACKGROUND_ELEMENTS}\nMain text: {MAIN_TEXT}\nSecondary text: {SECONDARY_TEXT}\nAccent symbol: {ACCENT_SYMBOL}\nWardrobe/styling: {WARDROBE_STYLE}\n\nKeep the upper field mostly empty. Place one large high-contrast athlete low and off-center. Use 6–12 long parallel horizontal bands behind the subject. Limit color to warm paper white, charcoal, and one accent color. Apply fine paper grain. No real teams, logos, or athlete likenesses.", "negative_prompt": "real team, real athlete, celebrity likeness, official logo, sponsor mark, jersey number, watermark, QR code, crowd, stadium, scoreboard, statistics, glossy 3D type, neon, gradients, smoke, lens flare, cartoon, painterly, excessive blur, malformed anatomy" }

The cleanest framing I’ve seen on this—separating loop closure, forward-rollout drift, and dual-latency is genuinely different from “make the video longer.” The open question your own critique points to: are these three actually orthogonal? A cache fix for consistency might trade against latency; the Error Bank might mask rather than reduce drift. An ablation per-mechanism would tell us whether it’s three solutions or one system pretending to be three.

Long-horizon world models hide three different problems. 1. Consistency is loop closure: leave and return, and the same place should persist. 2. Stability is forward rollout: self-generated errors should not compound. 3. Runtime requires both low visual and semantic latency. AlayaWorld assigns each a mechanism. 1. A rendered 3D cache retrieves by location; compressed frame history keeps recent dynamics. 2. An Error Bank perturbs memory and target during training, teaching the DiT to recover from imperfect history. 3. Four-step DMD cuts visual latency, while chunk-boundary prompt switching cuts semantic latency without regenerating past frames. This framing matters more than "longer video". But evaluation is qualitative: no loop-closure metric, drift curve, latency table, or ablation. Paper: arxiv.org/abs/2607.06291 Github: github.com/AlayaLab/Alaya…

Postcards from Italy --sref 3606551899 687624522 --v 8.1 --preview

@Just_sharon7 @freebeat_ai This shot looks really cool, is it a one-shot?

This entire music video came to life in realtime while the music was playing on @freebeat_ai The video walks through my entire creative process. Watching every cut, transition, and scene lock perfectly to the soundtrack in real time genuinely feels like the future of video creation. Give it a try: freebeat.ai/?via=sharon-ri… @Zarnab_with_Ai @snskritinaruka @noracoman7 #freebeatRealtime prompt France vs England – The Final Battle Duration: 50–60 seconds Scene 1 – The Arrival (0–6s) Crowd roars. Stadium lights ignite beneath the night sky. Narrator: "Tonight... history calls. Two football giants. One unforgettable battle." Scene 2 – France (6–14s) The French captain walks through the stadium tunnel. Every step echoes with confidence as blue lights illuminate his path. He looks up toward the pitch. French Captain: "We don't fear pressure. We become stronger because of it." Scene 3 – England (14–22s) The England captain emerges from the opposite tunnel. The Three Lions crest gleams under the floodlights as thousands chant. England Captain: "We fight for every tackle... every pass... every dream." Scene 4 – Kickoff (22–32s) The referee blows the whistle. The ball rolls. Quick cuts of explosive dribbles, fierce tackles, pinpoint passes, diving saves, and roaring supporters. Narrator: "Speed meets strength. Passion meets pride. Every second writes another chapter." Scene 5 – The Duel (32–44s) France launches a brilliant attack. England counters instantly. Both captains race toward the ball in slow motion before colliding in a dramatic shoulder challenge. French Captain: "For France!" England Captain: "For England!" Scene 6 – Glory (44–56s) The stadium erupts as the final moments unfold. Fans wave flags while fireworks illuminate the night. Players celebrate together, showing mutual respect after an unforgettable contest. Narrator: "Some matches crown champions... but legends are remembered forever." Final On-Screen Line: FRANCE 🇫🇷 VS ENGLAND 🏴 One Game. One Dream. One Legacy.

@egeberkina The stick figure turned into a fashionista!

I think I'm addicted to this

@mattworkman Backroom escape

Backrooms didn't shot on IMAX, now we know why...