Cosmos 3 Super

ImagePremium

NVIDIA's flagship photorealistic text-to-image model with optional agentic refinement — generates multiple candidate images per round and rewrites the prompt between rounds for stronger prompt adherence. Top-end physics-aware photorealism, structured-JSON prompt expansion supported.

Capabilities

Text to Image

Specifications

Aspect Ratios

1:116:99:164:33:4

Estimated Generation Time

~12s

API Usage

cURL

curl https://www.flixly.ai/api/v1/generate \
  -H "Authorization: Bearer flx_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "cosmos-3-super",
    "prompt": "A beautiful sunset over the ocean",
    "type": "TEXT_TO_IMAGE"
  }'

Get your API key from the Developer Portal.