英伟达Cosmos：世界基础模型ComfyUI官方支持

🌹大家好！欢迎来到破狼公众号。感谢大家的支持与鼓励。在AIGC探索道路上，我将与你一路同行。喜欢就星标关注破狼公众号或文末扫码加入交流群 ！

Cosmos世界模型简介

• • 项目主页：https://www.nvidia.com/en-us/ai/cosmos/

• • Github：https://github.com/NVIDIA/Cosmos

• • 技术论文：https://research.nvidia.com/labs/dir/cosmos1/

• • Cosmos - 1.0 - Diffusion - 7B - Text2World：给定文本描述，预测121帧的输出视频。地址：https://huggingface.co/nvidia/Cosmos-1.0-Diffusion-7B-Text2World

• • Cosmos - 1.0 - Diffusion - 14B - Text2World：同上7B - Text2World模型，为14B大模型。地址：https://huggingface.co/nvidia/Cosmos-1.0-Diffusion-14B-Text2World

• • Cosmos - 1.0 - Diffusion - 7B - Video2World：给定文本描述和作为第一帧的图像，预测接下来的120帧。地址：https://huggingface.co/nvidia/Cosmos-1.0-Diffusion-7B-Video2World

• • Cosmos - 1.0 - Diffusion - 14B - Video2World：同上7B - Video2World模型，为14B大模型。https://huggingface.co/nvidia/Cosmos-1.0-Diffusion-14B-Video2World

Cosmos ComfyUI体验

当前ComfyUI官已支持Cosmos的运行，因此仅需更新ComfyUI本体到最新版本即可，无需安装其他插件。（注：本文涉及模型可文末网盘获取）

• • Cosmos视频模型：下载模型并放置 /ComfyUI/models/unet 目录下。这里包含7B和14B两个版本，建议使用7B模型。下载地址：https://huggingface.co/mcmonkey/cosmos-1.0/tree/main

• • cosmos_cv8x8x8_1.0.safetensors - VAE模型：下载模型并放置 /ComfyUI/models/vae 目录下。下载地址：https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/tree/main/vae

• • oldt5_xxl_fp8_e4m3fn_scaled.safetensors Clip模型：下载模型并放置 /ComfyUI/models/clip 目录下。该模型并非Flux等模型使用的基础T5模型，需要独立下载模型。这里包含fp16和fp8两个版本，建议使用fp8版本。下载地址：https://huggingface.co/comfyanonymous/cosmos_1.0_text_encoder_and_VAE_ComfyUI/tree/main/text_encoders

Flux文生图&混元视频工作流

• F.1-绮梦流光-水湄凝香：

https://www.liblib.art/modelinfo/134c6dd95aef48e98a22b24e003e026b

• 文生图-Flux文生图(PuLID|LORA|Joy|SUPIR)工作流：

https://www.liblib.art/modelinfo/782aacd70f604da39e83368c696a02a8?versionUuid=9c5eceb01fb94d4d93d60fe2c0bd7468

• 文生视频-腾迅混元最强开源视频(LORA)工作流：

https://www.liblib.art/modelinfo/35ee21d5f6a94204abb767ad194ab9cd?versionUuid=be674032ffa14e5597a08922556f4da0

Cosmos世界模型工作流体验

• Cosmos世界模型的14B模型本地也可运行，至少需要大约需要24.8G显存（虚拟内存），848*480分辨率121帧（5秒视频）采样耗时大约需要14分14秒，总共耗时约17分。

01.仓库机器人

The video is a first-person perspective from the viewpoint of a large, humanoid robot navigating through a chemical plant. The robot is equipped with a camera mounted on its head, providing a view of the surroundings. The environment is industrial, with large metal structures and shelves filled with various boxes and supplies. The robot is seen moving forward, with its camera capturing the scene from a height of about 1 meter above the floor. The camera remains mostly static, with slight movements as the robot advances. The robot's body is metallic, with a large, boxy structure and a prominent head with a camera. The background is filled with industrial equipment and storage shelves, indicating a busy and functional workspace. The lighting is bright, typical of an industrial setting, with overhead lights illuminating the area. The robot's movement is steady and deliberate, suggesting a purposeful task, possibly involving inspection or maintenance. The video does not contain any text overlays or channel logos, focusing solely on the visual experience of the robot's journey through the plant.

02.城市自拍

The camera records horizontally a fashionable Chinese beauty at the Bund. Her face is beautiful and her skin is fair. Her black straight hair is over her shoulders. She is wearing a cool and ultra-thin light blue dress, and the hem of the dress flutters in the wind. The beauty holds her mobile phone and takes a selfie facing the Huangpu River intently, with the corners of her mouth raised. Behind her are the bustling buildings and bustling crowds at the Bund, with bright lights. The river wind gently blows, her hair slightly flutters, and her smile is brilliant. The picture is clear and the colors are bright.

03.机器人

A humanoid robot is showcased in a minimalistic setting, standing on a reflective black surface against a plain dark background. The robot is predominantly white with black accents, featuring a sleek, futuristic design. Its head is rounded with a smooth surface, and it has a pair of blue, triangular eyes that give it an expressive appearance. The robot's torso is compact, with a rectangular black panel in the center, possibly indicating a display or sensor area. Its arms are articulated, with visible joints at the shoulders, elbows, and wrists, allowing for a range of motion. The hands are designed with three fingers, suggesting a capability for basic grasping or gesturing. The robot's legs are sturdy, with visible joints at the hips, knees, and ankles, providing stability and mobility. Throughout the video, the robot remains stationary, maintaining a slightly bent posture with its right arm raised and left arm lowered, as if in a welcoming or waving gesture. The camera remains static, focusing on the robot, highlighting its design and features without any movement or zoom. The lighting is soft, casting subtle shadows that enhance the robot's contours and the reflective surface beneath it, creating a clean and modern aesthetic.

04.多机械臂协作

A sophisticated robotic assembly line features two advanced robotic arms working in tandem. The left arm, sleek and black, is equipped with a precision gripper, delicately handling a rectangular package labeled 'Kawada.' The right arm, robust and white, is similarly outfitted, poised over another package. Both arms are mounted on a metallic base, surrounded by a grid-like conveyor system. The background reveals a clean, industrial setting with a focus on automation and efficiency. Bright, even lighting highlights the mechanical details, casting subtle reflections on the metallic surfaces. The scene remains static, emphasizing the precision and coordination of the robotic arms as they perform their tasks with meticulous accuracy.

05.机械臂作业

An industrial robotic arm, painted in a vivid orange hue, dominates the foreground, positioned within a modern factory setting. The arm is intricately designed with multiple joints and a complex network of cables, showcasing advanced engineering. It is equipped with a specialized tool at its end, poised for precision work. The background features a spacious, well-lit industrial environment with concrete walls and a high ceiling, illuminated by fluorescent lights. Various machinery and equipment are scattered throughout, adding to the industrial ambiance. The scene is static, with the robotic arm as the focal point, emphasizing its technological sophistication and the meticulous organization of the workspace.

06.动车

A beautiful Chinese woman with long hair is sitting on a high-speed train, preparing to travel She was wearing a cool and slim white dress, very elegant The sunlight shone through the car window, illuminating her fair skin and gentle smile She quietly watched the scenery passing by outside the window, her hair swaying gently in the wind The camera captured her quiet and beautiful moments from the side The entire screen is presented with high-quality images, giving people a comfortable feeling

07.沙滩

The camera focuses horizontally on her.  Her skin is fair, and her features are delicate and extremely charming.  Her black big wavy long hair is scattered randomly on her shoulders.  The beauty is wearing a cool and ultra-thin purple strapless top.  The silk fabric shimmers with a soft light, highlighting her charming collarbone and fragrant shoulders.  She holds the mobile phone, turns slightly sideways facing the sea, showing a graceful curve, and the posture is sexy and alluring.  Around is the golden sandy beach.  The sunlight pours down, leaving charming light and shadow on her.  Her eyes are enchanting, and her red lips are slightly parted, as if whispering, exuding fatal attraction.

如果觉得文章不错，就请赞、在看与转发三连