falText to imageComing soon

nvidia/cosmos-3-super/text-to-image

Name: nvidia/cosmos-3-super/text-to-image
Brand: fal

nvidia/cosmos-3-super/text-to-image is a text to image tool from fal that runs in the background and returns images. Coming soon: listed with its price, not yet open to runs.

This tool is listed but not yet available to run.

Priced by usage

Pricing

$0.052 per image. Billed on num_images, default 1. The final charge comes from usage.

Limits

Runs: In the background
Returns: image
Longest queue wait: 3600s
Run deadline: 1800s
Largest inline input: 1024 KB
Largest input file: 500 MB
Largest output: 4768 MB
Outputs per run: 16

Input

Field	Kind	Default	Description
`prompt` required	text	—	Text prompt describing the image to generate.
`negative_prompt`	text		Content to steer the generation away from (colors, objects, artifacts).
`enable_prompt_expansion`	true or false	false	Expand the prompt with OpenRouter before image generation. When enabled, the prompt is rewritten into the dense structured-JSON format Cosmos3 was trained on; generation falls back to the raw prompt if expansion fails.
`enable_agentic_generation`	true or false	false	Automatically generate and compare multiple candidate images, then refine the prompt between rounds to better match the original request. This can improve prompt adherence but increases latency and billable image generations.
`agentic_max_iterations`	number (whole) 1 to 3	2	Maximum number of refinement rounds when agentic generation is enabled.
`agentic_samples_per_iteration`	number (whole) 1 to 3	2	Candidate images to generate and judge per agentic iteration. The best candidate advances to the next rewrite stage.
`agentic_early_stop`	true or false	true	Stop early when a candidate image is already a strong match for the prompt.
`image_size`	JSON	square_hd	The size of the generated image. Each edge is clamped to 512-1280px (multiples of 16).
`num_inference_steps`	number (whole) 1 to 50	28	Number of denoising steps. More steps yield higher quality but take longer.
`guidance_scale`	number 0 to 20	4	Classifier-free guidance scale. Higher values increase prompt adherence at the cost of diversity.
`num_images`	number (whole) 1 to 4	1	The number of images to generate.
`seed`	number (whole)	—	The same seed and prompt given to the same model version will produce the same image every time.
`output_format`	one of jpeg, png	jpeg	The format of the generated image.

Same request in code

curl https://api.spiral.ai/v1/runs \
  -H "Authorization: Bearer $SPIRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "provider": "fal",
    "model": "nvidia/cosmos-3-super/text-to-image",
    "input": {
      "prompt": "A photorealistic close-up of two damp hands shaping a spinning cylinder of wet gray clay on a pottery wheel, fingers pinching the walls upward into a…"
    }
  }'