Dive into the latest developments in the AI world. Leave an emote if you want more recaps like these.
Please open Telegram to view this post
VIEW IN TELEGRAM
❤1
This media is not supported in your browser
VIEW IN TELEGRAM
- Reference(s): Optional
- Styles Support: Realism (tested)
"15-second photorealistic miniature disaster film. A tiny human explorer only 3 centimeters tall stands on a cluttered breakfast table. Everything around him is gigantic. 0-3s: The explorer looks upward in terror as an enormous human hand lifts a coffee cup. 3-8s: Coffee spills across the table like a massive tidal wave. The tiny explorer sprints between gigantic cereal pieces, climbs a spoon and jumps onto a floating sugar packet. 8-12s: Cinematic low tracking shot as the sugar packet surfs the advancing coffee wave toward the edge of the table. 12-15s: The explorer leaps onto the safe surface of a giant toast slice, only to see an enormous butter knife descending from above. End on the shocked reaction. Realistic scale, convincing fluid physics, macro photography, energetic action, natural sound effects."(20% discount /w code "L1TMGOQ2" from all plans for 1st month
Please open Telegram to view this post
VIEW IN TELEGRAM
PRO PROMPTS (AI) | SEEDANCE | GROK | WAN | HAILUO |
This media is not supported in the widget
VIEW IN TELEGRAM
PRO PROMPTS (AI) | SEEDANCE | GROK | WAN | HAILUO |
So, when I hosted the contest, there was a massive bot wave of people coming in just for this. Hopefully, this follower was an OG fan at least and not a bot. We went from 1k followers to 24k and then now to 8k. Crazy stuff. 😂
😁1
This media is not supported in your browser
VIEW IN TELEGRAM
- Reference(s): None
- Styles Support: Realism (tested)
"15-second photorealistic surreal comedy, 9:16. A tired office worker in a gray suit stands inside a perfectly ordinary modern elevator, viewed from a stationary camera at the back. 0-4s: Elevator doors open, revealing a vast medieval battlefield with charging armored knights and a fire-breathing dragon flying overhead. The worker looks confused and presses the close button. 4-8s: Doors reopen onto a deep underwater civilization, enormous whales swimming past glowing skyscrapers beyond the elevator threshold. The worker closes the doors again. 8-13s: Doors open onto the surface of the moon, where a group of astronauts casually plays basketball beneath Earth. 13-15s: The worker silently sighs, presses the close button and looks directly at the camera. Consistent elevator interior, natural lighting transitions, convincing depth and scale, no dialogue."(20% discount /w code "L1TMGOQ2" from all plans for 1st month
Please open Telegram to view this post
VIEW IN TELEGRAM
Media is too big
VIEW IN TELEGRAM
Kimi K3 is most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window according to Kimi AI.
New model architecture: 2.5x the intelligence per unit of compute, not just more params.
Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.
Model weights: huggingface.co/moonshotai/Kim
Tech report: github.com/MoonshotAI/Kim
Tech blog: kimi.com/blog/kimi-k3
Please open Telegram to view this post
VIEW IN TELEGRAM
❤1🔥1
Media is too big
VIEW IN TELEGRAM
- Reference(s): Optional
- Styles Support: Anime (tested)
"2D动画动作镜头，高级现代热血动画美术，25 FPS。人物严格参考@Image1：年轻女性，短银白发，发尾渐变青蓝色，紫色高科技遮眼护目镜，黑紫色短款机能夹克，黑色贴身装甲上衣、露腰、高腰战术裤、腰带、露指手套、绑带护膝、黑紫短靴，服装有细窄紫色发光线路。脸型、发型、护目镜、身材、服装比例和颜色全程一致。她使用一把修长的黑紫色未来感长剑。极简黑暗空间，浅色地面，单一强顶光。赛璐璐平涂、2-3层硬阴影、强剪影。无对白。核心只有一次攻击：先把剑举过肩膀到头顶后上方蓄力，再从上到下完成一次强力纵劈。所有风格化MINI-CUT必须在剑真正开始下降后发生，全部是同一次纵劈的连续慢动作切片；绝不重置姿势、绝不变成第二次攻击。最终纵向斩击把画面本身切开进入纯黑。
镜头逻辑：前2.70秒必须极其克制，只建立人物、起手姿势和蓄力；禁止提前出现红色气场、速度线、墨刷、闪白、能量爆炸、环境变化或多余动作。真正的视觉爆发从剑开始向下移动的第一帧才启动。0:03.00之后的所有切换不是不同动作，而是同一段慢动作纵劈在不同“视觉处理层”中的连续推进；每次切换时，剑的位置必须比上一帧更低，肩、肘、腰、髋的旋转也必须继续前进，避免人物回到起手式。整体感觉像把约0.6秒的真实挥剑动作极度拉长，再用高密度mini-cut拆成数十个视觉瞬间。动作轨迹始终接近画面中央纵轴，由上向下，不做横扫、不旋转一圈、不跳跃。
能量逻辑：剑在起手时只是普通武器；只有下降后白色刃光才出现，并随下落逐步增强。白光始终贴合真实剑刃，不要变成脱离武器的随机光束。中段允许能量溢出成细碎紫红电弧、红色残影和白色纵斩痕，但所有效果都必须被剑的向下速度拖向上方，强化重量与方向感。斩击越接近地面，白色纵痕越粗、越亮，最终成为贯穿画面的“切口”。
人物表演：她始终冷静、自信，不喊叫、不夸张张嘴。护目镜遮住双眼，因此情绪主要通过下半张脸、下颌、身体张力和护目镜反光表现。银白到青蓝的短发在挥剑过程中受气流向上与向后掀起；夹克下摆、腰带挂件和裤腿细节跟随惯性延迟摆动。保持@Image1的女性体型与服装，不增加披风、裙子、盔甲、长发或其他饰品。
剪辑质感：mini-cut切换必须利落、硬切、几乎无转场；只有少数白闪可作为冲击桥接。慢动作感来自姿态连续推进、smear原画、残影和每个0.2-0.3秒切片，而不是让人物停住不动。声音只保留剑的低频蓄力声、空气被切开的尖锐whoosh、能量噼啪与最终画面撕裂声；无对白、无音乐。
0:00.00-0:01.05 正面全身。她在顶光下把剑举过肩膀，继续抬到头顶后上方，双手握柄，准备纵劈。正常全彩动画，无气场、无墨迹、无闪光。
0:01.05-0:02.05 中近景。剑到达最高起手位，肩背收紧，身体微微后压，重心稳定。镜头轻微低角度推近，只有蓄力，没有特效。
0:02.05-0:02.70 上半身特写。停顿半拍，紫色护目镜反射微光，双手握紧，肩膀开始向前下方发力。
0:02.70-0:03.00 纵劈正式开始。剑从头顶后上方向正前方下落，进入轻微慢动作；刀刃开始下降的瞬间，细窄白光第一次点亮剑锋。
0:03.00-0:03.22 MINI-CUT：剑刃极近特写，白色刃光向下拖成长线，粒子向上反甩。
0:03.22-0:03.44 MINI-CUT：护目镜与脸部特写，镜面映出下降中的白色剑光，表情冷静。
0:03.44-0:03.67 MINI-CUT：侧面半身，肩、手肘、腰同步带动剑继续向下。
0:03.67-0:03.90 MINI-CUT：米白纸面+纯黑人物剪影，剑化为自上而下的白色直线。
0:03.90-0:04.14 MINI-CUT：粗黑墨刷沿纵劈方向从顶部猛撕到底部。
0:04.14-0:04.38 MINI-CUT：粗糙铅笔原画帧，同一动作继续，剑已降到脸部高度。
0:04.38-0:04.62 MINI-CUT：密集漫画速度线全部由上向下爆发。
0:04.62-0:04.87 MINI-CUT：近乎纯黑，仅以白色轮廓描出人物、双臂和下降中的剑。
0:04.87-0:05.12 MINI-CUT：黑底红色剪影，剑为纯白，红色气场因下劈速度向上拖尾。
0:05.12-0:05.37 MINI-CUT：铅笔原画+红色粗糙能量裂纹，剑继续下降到胸前。
0:05.37-0:05.62 MINI-CUT：白底黑色人物，剑路周围出现纵向红边冲击环。
0:05.62-0:05.88 MINI-CUT：黑底红色身体，白剑向下加速，头发和夹克被气流向上掀起。
0:05.88-0:06.14 MINI-CUT：米白纸纹，黑墨像被刀锋自上而下割裂。
0:06.14-0:06.40 MINI-CUT：超密速度线+轻微画面震动，剑降到腰部附近。
0:06.40-0:06.67 MINI-CUT：双手握柄特写，手腕翻转控制剑刃保持垂直轨迹。
0:06.67-0:06.95 MINI-CUT：剑气特写，白色核心外包紫红细碎电弧，能量不稳定外溢。
0:06.95-0:07.24 MINI-CUT：红黑剪影，身体进入斩击中后段，肩和髋同步下沉。
0:07.24-0:07.54 MINI-CUT：单色原画帧，巨大白色纵向斩痕开始从顶部延伸。
0:07.54-0:07.85 MINI-CUT：瞬间全彩，剑锋从镜头前高速向下掠过，强烈近景透视。
0:07.85-0:08.17 MINI-CUT：黑场，仅剩白色人物轮廓与纵向剑痕，剑痕越来越亮。
0:08.17-0:08.50 MINI-CUT：高质量拉伸变形原画smear，手臂、剑、衣摆保持同一次下劈。
0:08.50-0:08.83 MINI-CUT：红黑画面，白色纵斩已贯穿画面大半。
0:08.83-0:09.12 MINI-CUT：纯白冲击帧，黑色碎裂线从剑路两侧爆开。
0:09.12-0:09.42 MINI-CUT：全彩侧面，剑接近最低点，头发和衣服被力量向上卷起。
0:09.42-0:09.72 MINI-CUT：护目镜特写，紫色镜面被一道垂直白光分成左右两半。
0:09.72-0:10.04 MINI-CUT：黑色剪影，巨大白色纵向剑弧从画面顶部直落到底部。
0:10.04-0:10.37 MINI-CUT：远景，明亮纵向剑气完整贯穿人物前方空间。
0:10.37-0:10.72 MINI-CUT：白闪、空间折射和短暂画面错位，像空气与镜头都被切开。
0:10.72-0:11.25 回到全彩中景，同一记纵劈终于落到底。她保持低位收势，剑尖接近地面，绝不追加第二次挥剑。
0:11.25-0:11.95 巨大白色纵向斩痕留在屏幕中央，从顶部贯穿到底部，边缘闪烁紫红细光与碎粒。
0:11.95-0:12.70 镜头停住。斩痕像真正切进影像的裂缝，左右画面开始轻微错位，裂缝内部为纯黑。
0:12.70-0:13.45 裂缝扩大，整幅画面沿这条自上而下的刀口分成左右两半，如真实被切开的胶片。
0:13.45-0:14.20 左右两半向外分离并被黑暗吞没，人物、地面、灯光一起被纵劈“切掉”。
0:14.20-0:15.07 纯黑保持，无文字，无额外镜头。."(20% discount /w code "L1TMGOQ2" from all plans for 1st month
Please open Telegram to view this post
VIEW IN TELEGRAM
> shooting-star-red-alpha
> shooting-star-green-alpha
This offers better physics, realism, and texture compared to the current GPT Image 2.5. There is a possibility that this could be the New Nano Banana model as well (no confirmation).
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
❤1
More about the Mystery "Iron Goddess" video model, likely to be Happyhorse 1.2 or Wan 4.0. After the disappointing release of Kling 4.0, I can see this being a close second to Seedance 2.0/2.5 right now.
https://x.com/TheoMediaAI/status/2114029689845223593
https://x.com/TheoMediaAI/status/2114029689845223593
X (formerly Twitter)
Theoretically Media (@TheoMediaAI) on X
New Mystery AI VIdeo Model: Who is Iron Goddess?
I have a guess here, but let me know what you think!
I have a guess here, but let me know what you think!
Media is too big
VIEW IN TELEGRAM
- Reference(s): Optional
- Styles Support: Realism
"16:9 宽屏，电影质感实拍风格，时长 15 秒，呈现一场节奏紧凑、连贯流畅的跑酷式格斗。镜头切换仅采用硬切（直接跳转），无叠化、变形或其他转场特效；每次切换时，动作与运动方向均保持连贯衔接。全片无对白，仅保留真实的脚步声、衣物摩擦声、水花飞溅声及撞击声。
主角：@image 1（面部特征、发型、玫瑰粉色挑染、侧边长辫、李子色上衣及米色工装裤，全片保持与参考图一致）。她的动作干脆利落、果断且流畅。她在整场格斗中占据绝对主导，以迅雷不及掩耳之势接连击倒十名蒙面袭击者。
对手：十名蒙面的“灰烬卫士”（Ashen Guard），身穿炭灰色连帽雨衣，面部裹着石板灰色的布巾，手持竹杖与短棍。他们从上方、侧面及后方包围而来。每一名对手在被击中后都有明确的物理反馈：或向后倒退、失去平衡，或被撞向墙壁、架子或栏杆，随后退出画面；紧接着，下一名袭击者便会补位冲上。
场景：雨后，位于陡峭多雾山坡上的一处废弃梯田茶园。此处布满长满苔藓的石砌梯田墙、狭窄湿滑的石阶；一条浅浅的灌溉水渠顺坡而下；还有断裂的木制人行桥、锈迹斑斑的茶叶晾晒架、倾斜的竹制脚手架，以及悬崖边缘的一道细长瀑布。低垂的金色阳光穿过薄雾，从侧面斜射进来。色调：苔藓绿、湿润的石板灰、暖琥珀色、深李子色。每一次撞击都会激起水花、泥点、破碎的茶叶和雾气。仅采用写实的自然光照明。
镜头 1（0-2.5 秒）：低机位，贴近地面，视场角（FOV）35 度。主角从石梯田底部猛然跃起，借台阶发力，一记飞身前踢正中第一名袭击者的胸口，将其撞向石墙。落地后，她侧身滑行，避开第二名袭击者挥来的竹杖，随即以一记上挑肘击中对方下巴；对方应声倒地，平拍在潮湿的茶丛之中。镜头 2（2.5-5秒）：侧向跟拍，视场角 40 度。她将脚踏在长满苔藓的梯田挡土墙上，向上奔跑三步，随即斜向踢向“袭击者 3”；对方失去平衡，从台阶上滚落下去。她动作不停，借一根斜靠的竹竿腾空而起，膝盖猛撞“袭击者 4”，致使其瘫倒在石面上。
镜头 3（5-7.5秒）：摄像机贴近地面掠过并向上扫视，视场角 30 度。她压低身姿，在悬垂的竹子和绳索下疾速下坡；随后旋身横踢，将“袭击者 5”踢得撞向一个生锈的茶架。她以手撑地翻滚，紧接着飞踢“袭击者 6”，使其向后滑去，撞在梯田挡土墙的边缘。
镜头 4（7.5-10秒）：摄像机绕她旋转拍摄，视场角 32 度。她双膝跪地，滑过浅浅的灌溉渠，激起大片水花，随后在坍塌的人行桥旁跃起。她跨越缺口，在半空中扭转身体，用肩膀猛撞“袭击者 7”，将其撞落至下方的梯田。落地后，她接住“袭击者 8”掷来的竹竿，撑地跃起，双脚同时踢向“袭击者 8”和“袭击者 9”，将两人向相反方向踢飞。
镜头 5（10-12.5秒）：摄像机先如追逐般向前推进，随后迅速后拉，视场角 38 度。她沿着倾斜的竹制脚手架奔跑，随后向后空翻，避开了“袭击者 10”从后方的偷袭。她在半空中扭转身体，一记后踢正中对方肩胛骨之间，使其向前滚落坡下。
镜头 6（12.5-15秒）：摄像机降至低位广角，视场角 30 度，随后锁定为平视角度的正面镜头。她落地后压低身姿，滑过一处水洼，激起四溅的水花；随即起身，摆出一个利落的战斗姿态。因高速动作，她的辫子和衣摆仍在摆动。动作序列在这一强劲、静止的姿态中戛然而止，最后一帧画面清晰锐利。
镜头：采用动态追踪拍摄，紧随其重心移动（涵盖跳跃、跑墙、撑杆跳、踢击等动作），而非固定不动。强调冲击感的摄影手法：每次击打时画面伴有轻微而干脆的震动；镜头会短暂跟随被击飞的对手，随后迅速切回至她身上。快速动作中呈现自然的动态模糊。
节奏：通过爆发性动作不断累积速度感。她的动作干脆果断；敌人的反应则迅速且沉重。全程无慢动作、无停顿、无刻意摆拍，绝无多余动作。
视觉品质：超高清画质，电影级质感；皮肤、织物及水体物理效果逼真，呈现出实拍影像的观感。
固定要素：粉色挑染保持柔和的玫瑰色调；辫子始终位于右肩；身着梅紫色上衣和米白色工装裤；靴子与手套样式固定；共有十名身穿炭灰色外套、面部缠绕石板灰色布条的袭击者。
限制条件：至少十名敌人，且全部被击中。无落空踢击；每一次攻击都必须命中目标。动作不重复。严禁出现奇幻光效、霓虹色彩或霓虹光晕；不得呈现卡通或CGI风格；肢体不得扭曲变形；严禁模型穿模（物体相互重叠穿透）；动作必须流畅无卡顿；画面中不得出现额外人物，亦不得包含任何文字或水印。"(20% discount /w code "L1TMGOQ2" from all plans)
Please open Telegram to view this post
VIEW IN TELEGRAM
❤1
Media is too big
VIEW IN TELEGRAM
- Reference(s): Optional
- Styles Support: Realism
"Create an ultra-photorealistic live-action Japanese supernatural action sequence with high-budget Hollywood cinematography, realistic physics, dark fantasy atmosphere, and practical-feeling VFX. A Japanese female shinobi with long black hair, wearing a fitted black ninja outfit, layered dark robe, forearm wraps, and traditional black boots, stands on a devastated battlefield surrounded by ruined shrines, broken stone structures, cracked ground, and swirling dust. HAND SEALS: She faces a powerful enemy and rapidly performs precise ninja hand seals. Strong winds intensify as the camera pushes toward her focused face. She completes the final seal and forcefully shouts in Japanese, 「死鬼封尽！」 (“SHIKI FŪJIN!”). A powerful shockwave blasts dust and debris outward. REAPER MANIFESTATION: A gigantic terrifying Shinigami materializes behind her, towering over the battlefield with a pale corpse-like face, terrifying supernatural eyes, long wild white hair, dark ancient Japanese robes, and enormous spectral arms. The camera circles around them, revealing its immense scale as wind and debris react naturally to its presence. SPIRITUAL CHAINS: The Shinigami extends its enormous hand, releasing thick supernatural chains that wrap around the enemy's torso, arms, and legs. The enemy struggles desperately as the chains drag them backward across the cracked battlefield, scattering stones and dust. SOUL EXTRACTION: The Reaper's spectral arm passes harmlessly through the female shinobi and reaches the enemy. Its enormous hand grabs the enemy's spiritual essence, forcibly extracting a translucent humanoid soul from the intact physical body. The soul stretches and struggles against the supernatural force while the chains hold the enemy in place. FINAL SEAL: The Shinigami pulls the soul into its open mouth and absorbs it. A massive shockwave erupts across the battlefield. The enemy collapses to their knees, completely immobilized. The woman releases her hand seal, remaining calm and fearless as the gigantic Reaper fades into the drifting dust behind her. End with a dramatic wide shot of the woman standing alone amid the ancient ruins. VISUAL & AUDIO: Ultra-realistic skin textures, natural hair movement, detailed fabric physics, cinematic lighting, volumetric dust, realistic supernatural effects, dynamic camera movements, and consistent character appearance. No music, narration, subtitles, or additional dialogue. Only natural battlefield sounds, powerful wind, supernatural rumbling, chains, soul extraction effects, and the clearly pronounced Japanese activation phrase. Avoid anime, cartoon visuals, cheap CGI, excessive neon glow, distorted anatomy, inconsistent faces, unrealistic physics, blood, and gore."(20% discount /w code "L1TMGOQ2" from all plans)
Please open Telegram to view this post
VIEW IN TELEGRAM
❤1
Media is too big
VIEW IN TELEGRAM
- Reference(s): Optional
- Styles Support: Realism
"Create a 30-second, photorealistic live-action miniature-person VFX comedy video showing an ordinary shower turning into an extreme survival adventure. Use cinematic macro photography, ultra-shallow depth of field, realistic water physics, dynamic handheld camera movements, dramatic lighting, and spectacular scale contrast. The video should feel like a premium Hollywood miniature-action sequence filmed inside a real bathroom, with hilarious near-disasters and visually satisfying moments. Ultra-detailed 4K quality, realistic textures, natural motion blur, and physically believable interactions. CHARACTER CONSISTENCY The main character is a tiny adult woman approximately 1:5 human scale, with realistic facial features, expressive eyes, shoulder-length dark hair tied in a small ponytail, a pastel-blue sleeveless top, dark shorts, and tiny white sneakers. Keep her face, hairstyle, clothing, body proportions, and scale perfectly consistent throughout the entire video. She must look like a real miniature person, not an animated or cartoon character. SCENE SEQUENCE Scene 1 — The Danger Begins A tiny woman stands on the edge of a bathroom floor beside a giant shower area, confidently preparing for a shower. The camera slowly pushes in as she looks up at the enormous showerhead above her. Suddenly, the shower turns on, and her confident expression changes to pure terror. Scene 2 — The First Waterfall A massive stream of water crashes down like a powerful waterfall. She runs for cover as giant droplets explode against the tiles around her. One enormous droplet splashes beside her, sending her tumbling backward across the wet floor. Scene 3 — Soap Bar Avalanche She climbs onto a giant soap bar to escape the rushing water. The slippery soap suddenly slides across the floor like a runaway skateboard. She desperately balances on top, waving her arms while racing toward the drain. Scene 4 — The Shampoo Bottle Collapse She jumps off the soap bar and grabs the edge of a giant shampoo bottle for safety. The bottle wobbles, tips over, and releases a thick stream of shampoo onto the floor. She slides down the slippery shampoo trail like a water slide, barely avoiding a giant puddle. Scene 5 — The Drain Escape The water begins flowing rapidly toward the bathroom drain. She runs in the opposite direction as tiny waves chase her feet. At the last moment, she leaps onto a raised tile edge and grabs a small grout line, dangling dramatically above the rushing water. Scene 6 — The Final Victory She pulls herself onto a dry bathmat, exhausted but triumphant. Behind her, the shower suddenly releases one final giant splash that sends her flying forward onto the soft mat. She slowly lifts her head, gives the showerhead an angry look, and shakes her fist at it. End with a dramatic close-up of her soaked, furious face. CAMERA AND VISUAL STYLE - Cinematic macro close-"(20% discount /w code "L1TMGOQ2" from all plans)
Please open Telegram to view this post
VIEW IN TELEGRAM
Our latest frontier model. Carbon delivers improved performance across domains — ranking #1 on v5.0 of the ArtificialAnlys Intelligence Index and sets a new state-of-the-art in cybersecurity and long-horizon capabilities.
It’s rolling out starting this week to cyber defenders through our Fairwind Program, with expanded access coming soon https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/?utm_source=tw&utm_medium=social&utm_campaign=og
Please open Telegram to view this post
VIEW IN TELEGRAM
Please open Telegram to view this post
VIEW IN TELEGRAM
👍1