HedraImage to imageComing soon

krea-2

Name: krea-2
Brand: Hedra

krea-2 is an image to image tool from Hedra 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.

Price pending

Pricing

Price pending.

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	—	Generation prompt. From 1 to 5000 characters.
`aspect_ratio` required	one of 1:1, 4:3, 3:2, 16:9, 2.35:1, 4:5, 2:3, 9:16	—	Output aspect ratio.
`seed`	number (whole)	—	Seed passed to Krea. It does not make a render repeatable — Krea returns a different image each time from the same seed.
`images`	JSON	—	Reference images whose style — palette, lighting, texture, art direction — the generated image takes on. A reference with no clear subject of its own can crowd out the one you prompted for; lower `style_strength` to hold it back. Omit them to render from the prompt alone. 1 to 10 images, each at most 30 MB.
`quality`	one of medium, large	medium	Quality level to generate at. `medium` — the smaller model, strongest on illustration, anime and painting. `large` — the larger model, for photorealism and the effects that sell it — motion blur, grain, low dynamic range.
`creativity`	one of raw, low, medium, high	—	How loosely the model may read the prompt. 'raw' stays closest to it; 'high' drifts furthest for a more inventive result. Omit for Krea's own default ('medium').
`num_outputs`	one of 1	1	Number of outputs generated per job. Only 1 is supported.
`enhance_prompt`	true or false	false	Rewrite the prompt before generation. An LLM expands it into a fuller description and the model receives that text instead of the submitted one; the result's `prompt` reports what ran.
`style_strength`	number 0 to 1	—	How strongly the reference images pull the render toward their style, applied to every one of them. Krea pulls as hard as it can unless you say otherwise; lower this to leave more of the prompt's own subject, and 0 to ignore the references' style entirely. From 0 to 1.

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": "hedra",
    "model": "krea-2",
    "input": {
      "prompt": "<prompt>",
      "aspect_ratio": "1:1"
    }
  }'