spencer mckee

@its_unks hadn't read this before, great persistence breakdown

Remains one of my favorite pg essays: paulgraham.com/persistence.ht…

@JSFILMZ0412 Do you find you don't need as many retries?

Seedance 2.0 1080p just dropped 🚀 I put it head-to-head with OG 720p from early February… the difference is wild 👀 #Seedance2 #TopViewAi

@mangoster Process

Just recorded a full breakdown of my AI B-Roll process in this video i cover: - what i use to prompt each scene - fully trasnparent look at my iteration process - different style keywords (ready to be copy & pasted) - the trick to make AI footage look real comment 'PROCESS' + RT and i'll send it over (must be following so i can dm)

@DylanTFWang interested to see the editing workflow. marble seems to be lacking there

Genie3 generates videos. We generate 𝟯𝗗 𝘄𝗼𝗿𝗹𝗱𝘀 you can actually use. Launching tomorrow — Tencent #HYWorld 2.0, an engine-ready World Model🚀 This isn't a video. It's a real 3D scene, all generated & editable. One image in. A whole 3D world out. 🔥Open-source tomorrow

@LemonSliceAI this is awesome, congrats team!!

Introducing LemonSlice-2.1: the most advanced interactive avatars ever built Make your avatars do anything in real-time… wave, get angry, dance and more! Try the FREE research preview NOW And build avatars into your products using our API First 100 devs get $100 in credits 👇

@JSFILMZ0412 legend!

@spenmcke all seedance. im actually not happy with the diner. cindys voice changed a little bit

My recent Seedance 2.0 Short Film teaser version for the algo.

@theworldlabs exciting!

We're excited to be rolling out two model updates today! Marble 1.1: Improves lighting and contrast, with a major reduction in visual artifacts. Marble 1.1-Plus: Our new model built for scale. Create larger, more complex environments than ever before.

HappyHorse ranked higher than Seedance 2.0 on the eval? If this is Veo 4, it's game over for US video models

@john87445528 Nice. What's the performance like for long sequences?

找了个风格参考提示词给大家拿去做视频风格模拟用。【复制以下内容给 AI】： 角色设定： 你是一位拥有20年经验的电影摄影指导（DOP）和美术指导。你擅长从视觉画面中剥离内容，提炼出核心的美学逻辑。 任务目标： 请分析我上传的这组参考图，忽略画面中的具体人物和动作（这是可变量），只从以下维度深度解析画面中的“不可变视觉条件” (Immutable Visual Conditions)： 1. 光影逻辑 (Lighting & Atmosphere)： 它是硬光还是柔光？光源的方向性如何（侧光/顶光/背光）？明暗对比度（Chiaroscuro）是高还是低？是否有丁达尔效应或特定的环境光？ 2. 胶片与质感 (Texture & Fidelity)： 模拟的是什么年代的胶片（如70年代Technicolor）？颗粒感（Grain）的强度如何？画面的清晰度、锐度以及是否有物理损伤感（划痕、灰尘）？ 3. 色彩科学 (Color Grading)： 整体色调倾向是什么（如低饱和、高反差、偏绿/偏黄）？高光和阴影的色偏如何？ 4. 构图与空间 (Composition & Spatial)： 镜头语言有什么特征（如广角、长焦压缩）？前景遮挡关系如何？空间透视是平面的还是深邃的？ 输出要求： 请不要给我写散文。请直接提取出 10-15 个精准的英文关键词或短语（Prompt Tags），用逗号分隔，按重要性排序。这些词必须能准确复刻这种视觉风格。 二、将风格总结为模版的提示词 【标准化模版构建】提示词 （用来把提取的条件变成“万能模版”） 【使用场景】： 当你已经得到了第一步分析出的“风格关键词”（比如：Technicolor, heavy grain, hard light...），想把它变成一个以后能无限填空的“万能公式”时，把那些关键词发给 AI，然后发送这段话。 【复制以下内容给 AI】： 【复制以下内容给 AI】： 角色设定： 你是一位精通 Midjourney/Stable Diffusion 的 AIGC 提示词工程师。 任务目标： 基于我们刚刚提取的“不可变视觉风格关键词”，请为我构建一套“标准化、模块化的提示词模版” (Modular Prompt Template)。 模版结构要求： 请将提示词拆解为三个模块，并用括号 [] 标出用户需要填空的地方： 1. 【可变内容区】 (Variable Content Block)： - 留出 [景别 Shot Type] 的填空位。 - 留出 [主体与动作 Subject & Action] 的填空位。 - 留出 [环境描述 Environment] 的填空位。 2. 【半固定构图区】 (Semi-fixed Composition Block)： - 将刚才分析出的构图逻辑（如：前景遮挡、人小景大、垂直线条）写死在这里，作为固定后缀。 3. 【不可变风格区】 (Immutable Style Block) - 权重最高： - 将刚才分析出的光影、胶片质感、色彩关键词全部整合在这里。 - 关键： 必须包含负面提示词（Negative Prompts）建议，明确告诉 AI 不要生成什么（如：不要现代感、不要柔光、不要高清数码味）。 输出格式示例： 请最后直接给我一段可以复制使用的英文 Prompt 模版，类似于： [SHOT TYPE], [SUBJECT ACTION], [ENVIRONMENT]. [COMPOSITION RULES]. [IMMUTABLE STYLE KEYWORDS]. --no [NEGATIVE KEYWORDS] 案例在跟帖中。可以把这种方法用来复刻赛博胡金铨电影试试

@Artedeingenio Love the style. What's your workflow?

Tell me this isn’t magic, being able to create this kind of animation from a single Midjourney image. I’ll say it again: Midjourney + Seedance 2.0 is like having an animation studio working for you 24/7.

@yolandaaacao built in filter

just learned seedance 2.0 performs way better when prompts are chinese not claude/gpt chinese btw

@OdinLovis @fal Interested

4 billions years - part 1 , made in 2hours with Seedance 2.0 on @fal using api . Semi-automatic process vibe-coded. Monday I will share the full video and a tutorial on how to make this yourself ! Look at the end if you want to see Dinosaur ! I will show also how to do calibration of lighting etc..

@aisearchio What size?

Finally! Bytedance drops an image generator & editor that can run locally on your phone! Creates 1024x1024 images in only 3 seconds Will be open sourced soon! carlofkl.github.io/dreamlite/

@RadianceFields Clean

World Labs just dropped their newest model of Marble, Marble 1.1.

@Skywork_ai 40 fps is impressive

Skywork Matrix-Game 3.0 is here! FULLY OPEN SOURCE! Real-Time and Streaming Interactive World Model with Long-Horizon Memory - Fully open source: code, model, and technical report - 720p @ 40FPS with a 5B model - Minute-long memory consistency - Trained on Unreal Engine + AAA games + real-world data - Scales up to 28B MoE for quality, dynamics, and generalization Homepage 👉 matrix-game-v3.github.io Code 👉 github.com/SkyworkAI/Matr… Model 👉 huggingface.co/Skywork/Matrix… Tech report 👉 github.com/SkyworkAI/Matr… Create. Explore. Play. With Matrix-Game 3.0

@spaceandtech_ Seems to be sped up

A robotics startup in Shenzhen called Mind On has upgraded the Unitree G1 humanoid with an advanced robot brain. With this upgrade, the robot can perform everyday tasks on its own without human control. It was shown watering plants, opening curtains, cleaning, and moving items independently.

Should rename themselves to OpanClaud

@jang_yoel Just submitted! Excited to try it out

Excited to share these "wow" moments with the community 🤗. There's something special about watching a robot do something it was never shown how to do. Do apply to our public API to get access to run DreamZero without any GPUs!

2024 was the year of voice ai 2025 was the year of agents 2026 is the year of ____?