Try Google Gemini Omni Flash Video Generator from Google →
Models
Agents
WorkflowsStudioPricingBlogDocs
ExploreDiscover models by categoryBrowse All ModelsBrowse the complete catalogSee FavoritesSign in to view saved models
OverviewThe platform at a glanceLearnSkills, knowledge, guardrailsAnatomyWhat makes agents reasonBuild Your AgentPick skills, set tier, deploy
Pre-built AgentsBrowse the catalog
Agent Usecases
Ad Campaign ManagerApp Event ManagerApp Review RepliesBarber BookingCustomer Win-BackEcommerce ListingsRestaurant Reviews
Sign InStart Building

Task History

Click to see output list

No tasks yet

Go to Models
Explore models/
Video GenerationActive

nvidia / Cosmos3-Super

Cosmos3-Super

bynvidia

Cosmos 3 Super turns a reference image and motion prompt into a physics-grounded MP4 clip, with controls for aspect ratio, FPS, audio, and safety.

Text to VideoImage to VideoFast InferenceH200
Model ID
Cosmos3-Super
Provider
nvidia
Updated
1780907840
13
Comments
Average rating : 4.5 (15 users)
Providernvidia
ModelCosmos3-Super
Text to VideoImage to VideoFast InferenceH200
wiro playground—nvidia/Cosmos3-Super
Reset to defaults
0 / 1
Maximum 1 image allowed
Drop image to upload

OR

Click to browse your device

Supports: JPG, JPEG, PNG, GIF, WEBP, HEIC

Optional init image.

Required: Description of the desired animation, motion and camera movement

Required: Description of the desired animation, motion and camera movement

Sample outputs
Updated 1780907840
## Overview Cosmos 3 Super is NVIDIA’s 64B omnimodal world model, tuned for high-quality generation. It pairs a transformer “reasoner” with a diffusion “generator” to plan motion and render frames. For image-to-video, you give a first frame plus a motion description, and it synthesizes a coherent MP4 clip. This helps you prototype physically plausible motion for ads, visual tests, or physical AI datasets. ## What you can build - Animate product photos into short commercial-style clips with controlled camera movement - Create synthetic driving or robotics scene variations from a single frame - Generate motion studies for storyboards, previs, and concept trailers - Produce training clips for video analytics, tracking, and event understanding - Make looping “alive photo” assets for landing pages and app onboarding ## Inputs - A single starting image to use as the first frame (JPG, JPEG, PNG, WEBP, and sometimes GIF or AVIF). Use RGB images for best results. - A text description of the scene plus the motion you want. Describe camera moves, subject motion, and timing. - An optional negative description that lists artifacts or motions you don’t want. - A toggle to expand your prompt using a vision-language prompt rewriter that looks at the first frame. - A resolution choice for the output video (the UI may offer multiple tiers). - An aspect ratio choice for the output video (square, landscape, or portrait). - A frames-per-second setting to control playback speed and motion feel. - An optional seed value to repeat a result with the same inputs. - An advanced motion setting that shifts the sampler’s flow behavior. This can change motion stability and texture drift. - A toggle to add sound to the generated clip when supported by the run. - A toggle to enable or skip automated safety checks on the prompt and generated content. ## Outputs - An MP4 video file that starts from the conditioning frame and continues with generated frames. - The video output typically includes basic metadata such as width, height, FPS, duration, and frame count. - If sound is enabled and supported, the MP4 can include an AAC stereo audio track at 48 kHz, muxed into the same file. - The run also returns the seed used for generation so you can reproduce the same clip later. ## Limitations - Motion can wobble or drift across frames. Faces, hands, and fine text can change. - Fast camera moves can cause jitter, flicker, or object deformation. - Physics is learned, not simulated. Collisions and contact can look wrong. - Audio-video sync can be imperfect when sound is enabled. - Low-quality inputs reduce quality. Scanned images, heavy compression, watermarks, and motion-blurred photos often produce artifacts. - Grayscale images are not a good fit. Use color RGB inputs. ## Safety & compliance - Safety checks can filter or block certain prompts and outputs. - Keep safety checks on for public-facing or user-generated workflows. - Don’t use the model to generate disallowed content, illegal content, or deceptive media. - Cosmos 3 model weights and code are released under the OpenMDW 1.1 license. Review license terms before redistribution or fine-tuning.

Example prompts

Great starting points for Cosmos3-Super.

Cinematic, low-angle tracking shot of a vibrant neon green sports car. The camera smoothly glides backwards along the side profile, keeping the intricately designed black rear wheel in sharp focus. The wheel begins to slowly rotate, gripping the textured asphalt. Crisp, dynamic reflections of the surrounding environment and golden sunlight warp and slide across the highly polished green bodywork. In the out-of-focus background, the soft bokeh of trees sways gently in a light breeze, and the blurred figure shifts their weight slightly. High-end automotive commercial aesthetic, photorealistic, smooth and steady motion.Video Generation
High-angle drone shot gliding slowly forward over the historic Porto hillside. Cars move smoothly along the winding asphalt road, with a few vehicles disappearing into the tunnel while others drive past the crosswalk. Pedestrians are seen walking up and down the steep stone staircases and along the sidewalks. The warm, golden hour sunlight casts long, soft shadows across the terracotta rooftops and white building facades, creating a hazy, cinematic atmosphere. A gentle breeze subtly sways the green ivy covering the stone walls and the leaves of the scattered trees. In the distance, a few seagulls fly lazily across the sky above the large palace on the hill. The movement is fluid and realistic, capturing the vibrant life of the city.Video Generation
Close-up, highly detailed shot of a vibrant yellow bird with black and white striped wings perched on a curved black wire against a bright, overcast white background. The bird exhibits quick, hyper-realistic micro-movements, abruptly tilting its head to look around and blinking its dark eye. A gentle breeze subtly ruffles the fine yellow feathers on its chest and back. The black wire sways and bounces very gently as the bird shifts its weight. Photorealistic nature documentary style, sharp focus, smooth motion.Video Generation
Cinematic street-level shot. The rider on the black motorcycle drives smoothly and realistically from right to left across the street in the foreground. The leaves on the two round, green trees rustle gently in a light, natural breeze. In the bright blue sky above the historic, weathered green building, fluffy white clouds slowly drift across the frame. On the far left, the pedestrian subtly moves toward the shop entrance. The bright sunlight casts sharp, consistent shadows. Photorealistic, steady camera, natural everyday city motion.Video Generation

API quick start

Run Cosmos3-Super with a single API call.

POST https://api.wiro.ai/v1/Run/nvidia/Cosmos3-Super
{
  "prompt": "Slow camera pan-up revealing the Ballon d…",
  "inputImage": "https://your-cdn.com/input.png",
  "negativePrompt": "static animation, sudden cuts, flickering…",
  "width": 480
}
View full API docs

Discover, test, and run AI models, build workflows and agents with one unified API.

All systems operational
WiroAboutBlogCareersContact
ProductModelsAgentsPricingPartnerChangelogStatusFAQ
Getting StartedIntroductionAuthenticationProjectsCode ExamplesWiro MCP ServerSelf-Hosted MCPn8n IntegrationLLMs.txt
API ReferenceModelsRun a ModelModel ParametersTasksLLM & Chat StreamingWebSocketRealtime VoiceFiles
© 2026 Wiro AI. All rights reserved.
PrivacyTermsData Deletion